SEO John

2026年3月8日 下午 1:05|3小时 39分钟 36秒

关键词:
插件、分享、品牌、关键词、飞书、信息源、more languages、other languages、红人营销、竞争对手、用户体验、下载安装、more content、飞书用户、用户评论、content marketing、good thing、own language、other thing、other ones、own content、best content、real content、other popular languages、more platforms、other product companies、brand good、other English

文字记录:
说话人 1 00:00 
数百万英镑的营销预算，这里大家提到英镑，大家也不知道他来自哪里了，好吧，他今天分享的主题是如何构建一套以营收为中心的 Saas 增长引擎，跳出单纯的增长黑客就是 growth hacking 的套路。好，大家掌声欢迎来自英国 we scale STARTUP 的创始人 Daniel Johnson。

说话人 2 00:24 
谢谢您。

说话人 2 00:30 
大家好，我不说中文。

说话人 2 00:37 
So I hope you guys speak English well enough. I'm gonna try and keep slow so that everything makes sense. Is that cool? Cool. Yeah, cool. Awesome. I really like audience engagement. So if you have any questions, if you have any comments, if you disagree with me, feel free to shout, ask me questions. Go for it. So good. Awesome. Okay, so it's quite an honor to be the first talk of the day and I'm really glad to be here.

说话人 2 01:04 
Thank you so much for that opportunity. No, I started SEO years and years ago, not because I was like, I wanna be a marketer or even I'm particularly interested in SEO. I started it because I wanted a carbon fiber bike. I was really into cycling and I really enjoyed it. And I was doing all sorts of chores. I was like 14,15. And I would say to my parents, I would cut the grass, I would wash the car, I would clean the house. Just give me some money so I could buy this bicycle. But it was not enough. It was like 5 pounds here, 10 pounds there, and my bike was 1,000 pounds. So I realized I needed to do something else to figure out how do I make enough money to buy a thousand pound bicycle, right? And at the time, I was a bit of a .

说话人 2 01:54 
I really enjoyed playing with computers. I enjoyed breaking them down. I enjoyed putting them back together. I kind of understood how they worked. And it was at that time when I was starting a bunch of different businesses and I couldn't afford marketing for any of them, right? It was too expensive. So I had to learn how to do it myself. And that's kind of when I came across Google.

说话人 2 02:16 
Now, at the time, everything was really easy, was fantastic. If I had more keywords on my website than anyone else for that specific term, I would rank No. 1.

说话人 2 02:27 
Fortunately, things have advanced a lot more since then. It's all about user intent. And from that keyword frequency to where we are today has progressed so much. But when you go on live, you see so many issues. You see so much bad information about what SEO is and how it works. And what I wanna share with you today is a really simple framework that you can use to actually produce content that drives traffic, that produces revenue. Sound good? Fantastic.

说话人 2 03:00 
So most of my work has been involved in app, Saas and ECOM. So mobile application software service at events. And that's mainly because you have lots of data and you can predict stuff over time, which is why I really like those industries.

说话人 2 03:18 
I'm curious, so from 2025, who's produced more content that year than they did in 2024? Drinking? Well, okay, we got cool. Don't worry. I'm not gonna challenge you. If we can adopt, that's cool. So we've got four people, which is more content in 2025 than 2024. Interesting. So for those four people, I've got another question for you. Did you make more money? Yeah, we got one. Yes, I know that.

说话人 2 03:52 
Interesting. So I don't know if you're like me, but when AIS first came out, I was so excited about it. I produced so much content. I updated it to our website. And he did absolutely love it. He didn't drive any more revenue. It was kind of , honestly. And that was what I realized. I can't just copy and paste this into AI and hope that it's gonna work. So what I realize is that it's not about the volume of content that you produce, it's not even about necessarily the traffic that you drive, it's about the quality. And in order to produce high quality work, you really need to ensure that you have a good system.

说话人 2 04:35 
Now it could be a system, you can  framework, whatever it is to you, but having one in place is really important. Personally, I leave, I'm very ADHD. I want to produce everything right now. I want it live yesterday. I don't have to use the system because that's the restriction. But I've come to realize that it is really important in order to actually achieve the results that you're looking for.

说话人 2 04:59 
Who is assistant for content production process at the moment? Cool, then this is very valuable. Okay, so what I'm gonna cover is three items in the next few minutes. So the first is a system that you can use that would enable you to produce content at scale. Next is looking at the individual layers of that system, then figuring out how we gonna use AI as part of the process. And then finally, what can you do on Monday that will enable you to drive more traffic for website? So that sounds good. Excellent.

说话人 2 05:37 
The enthusiasm in this room is just driven. Okay, so on a very high level, the system that I usually use is ICP, intent map, activation layer, and revenue feedback. And.

说话人 2 05:55 
I think one of the interesting things is when I first started in marketing, I was really excited about the sexy bits. I love running ads. I loved the technical bits. I enjoyed all the exciting aspects of marketing. And you know what happened? Absolutely nothing. The first 10 companies that I started completely failed. And that was really depressing cuz I had no friends. I was at school. I, my grades are terrible because I was spending all of my time building the companies. And when you're building your own company, it is your baby. You're emotionally invested in the success or failure of that company. And I would do the same things again and again.

说话人 2 06:37 
I was like, okay, I need to work harder. I need to produce more content. And I realize that's not how it works. That's not how you should think about content production. And it's only after listening to someone else that I understand this.

说话人 2 06:50 
And it's interesting. Who knows IBM? I mean, yeah, okay. It is SAP. Okay, if you don't know SAP, congratulations. It is not a boring, difficult company on this place, right? It's an enterprise ERP platform. So at the time I was working with IBM as a consultant and they were trying to take over or trying to compete safety a bit more. And specifically it's Ivy Watson, which was their AI consuming all. And we were running a lot of content that was trying to bring people from SAP over to IDM. And we were saying to them, you know what? You can be 0.04% more efficient and impactful. Or if you use ideal lots, two cats, about zero points, 4%,0.04%, right? It means nothing.

说话人 2 07:44 
There's no, I don't look at that number and say to myself, yes, let's buy this million pound piece of software. Cuz that's not what happens. So what I did is I picked up the phone and I pretended to be a customer for these sat salespeople. We were pitching the product to. And what I would do is I would say to them, hey, I'm really interested. Tell me about how to use this. You know, tell me all the technical information. And they were talking through the software and I was like, yeah, I think that's boring. And at the end of a lot of these calls, they would say to me, I gotta go, I gotta pick up my kid from school. And I was like, okay, have a nice day. Goodbye, Saijin. And I found that really interesting. So based on those conversations, we then ran a bunch more contents. And from that process, we have two categories. We had informative and then we had to be emotional. The informative was 0.04% more efficient with the ideal ones compared to sap. Boring is held. I hate it.

说话人 2 08:50 
And then the other side of it was get home to your kids sooner. We might be able to. We don't know. Okay. Okay. So I said to them, if you use our software, instead of wasting your time on 0.4% impact, whatever it is, you can do your job more effectively. You can make your boss happier, and you can actually leave work on time because our software is better. And I knew you could do this for consumer products. I always knew that you could, you know, you go out and say, buy this protein powder. Thank you. Bigger muscles make you sexier, make you cooler. Okay, but I didn't realize you could do it for B to B. But at the end of the day, I think every individual, every human is, we make decisions based on emotion and we justify the logic. And it took me a long time to appreciate this because I read all the books and I didn't understand actually that emotions are really how we think. You know, we're not robots, something, Alice, this is kind of like the core information. And I think this is the biggest impact, the biggest thing that I Learned. It's not about the channels you use. It's about what are you using those channels for. So when applying this to SEO, what I'm gonna say here is like a lot of people will say best series of them or efficiency, you know, you don't wanna focus on things that mean nothing. Cuz if you see this, you're gonna be like, whatever. I see that so many times, doesn't make any time to me. But if you can speak to people and you can really understand what is their goal, what are they trying to achieve from an emotional level and from a specific level that forms the foundation of everything else that you wanna do, the better that you understand your customer, the more effective everything else would be. And I think. It's really difficult because a lot of us, we're like, yeah, I understand my customer.

说话人 2 10:58 
I know you're selling to, I understand the value that you provide, but actually getting on a fire with them, I think is the most valuable aspect. I was running marketing for the Saas company in London a few years ago, and I would take my clients out from there, will keep on YouTube. And I remember taking this guy out and we were at a pub and I said to him after a few bytes, and I said to him, how do you describe our software? And he said to me, it's a shop in one line of code. And I thought that was really interesting because I stumbled back to my office after 4 points and I updated some of the texts on our Google ads. And it then doubles in conversion rate. Because how we were just driving ourselves and how the client thought they would thought about our product, but two completely separate things. And I thought I understood this completely, but actually the company was a startup. Everyone else in the company was a developer apart from myself. And everyone was telling me, Daniel, we've got the fastest code on the internet. We, this is the cleanest Javascript you'll ever see. So when I was talking about the company, we're talking about the technical efficiency, the load time, response time, the frameworks will be using. But my client didn't actually care about that. What he cared about was that he could copy and paste a link into his website and instantly create a shop.

说话人 2 12:26 
And I think the first stage and most important stage is to really think from a such principal's perspective, what is the problem we're solving? And do we have a really clear understanding of who we're solving that for? Does that make sense? Yeah, awesome. Because I think we all know that. But at the same time, it's one of those things where it's only after like speaking with people, you really understand that alignment and how portion it is proof. So assuming you have a really good understanding of your customer, you can then understand the emotional problems they solve. And obviously you can create content calendar. I'm not gonna go through that today. And I'm sure that we on the speakers that will. Oh, oh, then. Okay. So the next layer is something that I think a lot of people really struggle with, which is actually like, hey, I've produced content, I've published, it's kind of ranking and then we get zero leads from that. What's going on. And I think the next focus there is the activation layer. There was a blog I was working on for one of my clients literally last week and was getting tons of traffic. And then we realized actually we didn't have a CTA embedded within the blog post itself. We went from zero leads to 15 leads in one week, which was fantastic. And it's something that we know we should do. But I think it's something also that kind of struggle with sometimes cuz it's something that we don't necessarily feel is going to offer an immediate value. So if it can match CTAs, I think that's one of the most important things, not just creating a generic CTA requester demo, but creating something that matches the content. Whether that's aligning the copy of the CTA, whether that's creating a dedicated type of CTA, or a lead minus.

说话人 2 14:18 
Next is actually making sure that any of the lead magnets you can choose or not a magic solution for everyone, they should really be solving a specific problem for your customer type. If you don't necessarily know what that is, it means you don't understand your customer effectively enough. And trial numbers, I think it's fairly obvious.

说话人 2 14:39 
And then retargeting loops. So who runs retargeting on their content?  melon. Okay. Run retargeting. And then when you make lots of money, message me and be like, Daniel, that was a really good idea. I wanna hire you. I'd like, okay, cool.

说话人 2 15:00 
So I think the other aspect is a lot of picking startups, a lot of the companies that I work with, there is a core relationship between marketing and sales because what happens is a lot of marketing and a lot of sales often get their data from two desperate, two different sources. I'm actually having this at the moment with one of my clients. We've got a sales lady and she looks at one source of information, and I try and tell her to look at another source of information and then she says no. And we continue on the same path again and again. And the problem with that is it results in us making different decisions every day, and they get so frustrated.

说话人 2 15:44 
And then there's also the conversation about friction through market and sales pick up the course for qualify lead, but hey, I wanna buy your product. The and then the sales department complains about things are good enough. So that's another story. So creating some sort of focus between marketing and sales, I think is really important. Sales teams informs content priorities. This is something that seems really obvious, but a lot of people don't do it. Get the transcripts of the sales conversations, run them through an AI like Claude, and then use that to identify what are the key jobs to be done and emotional challenges that people are solving. You can even automate that process.

说话人 2 16:27 
So what I do at the moment is with one of my clients, every time at the beginning of every week. It goes through all the transcripts of the sales conversations and then produces a report and then publishes that to Slack. And that says what were the biggest challenges we had? What were the biggest challenges that the client had? What were the words they use? And what was the emotional problem page.

说话人 2 16:49 
I also use AI to then update our overall messaging map so that every week, the way that we articulate our value proposition towards our customer becomes more and more specific, targeted and aligned. And the other thing that I think about is convention days refining ICPs. What I mean by that is kind of like we'll have these customer episodes, people that we believe to be our main target user. And for startups, a lot of people struggle with that because they say to themselves, well, No. 1, our customer is everyone, which is the biggest issue in marketing. Cuz if your customers everyone, your customer is no one.

说话人 2 17:31 
And then the second part is, because it's their baby, because it's their company, they're really worried about creating a custom. Some of the sooner that's wrong. There's this emotional friction to doing that. What I say to that is like develop a draft customer persona. And it's not about accuracy. It's just about creating something that works. And then every week have a conversation with your team about how do we make this more accurate, how do we make it more specific? Because it's not like a static document. It's a document that evolves every time. And I think it's really important to use the information that we have, every customer conversation to be able to improve the targeting and to make that more refined as we progress. And then going to customer insights, feel new content. So using those emotional jobs to be done to understand how to drive more in QLS. Does anyone use AI at the moment to use AI with their sales conversations to produce content? Interesting. I'm kind of curious, how do you use, is it the sales transcripts or is it something else?

说话人 3 18:40 
I use SEO bot to produce pages like blogs.

说话人 2 18:47 
And do you use the sales conversations as well?

说话人 3 18:51 
I heard about one like recently, he's quite famous. Haven't use it yet. Think about use it.

说话人 2 18:59 
Nowadays it's so easy. Like I highly recommend it if you're not using it already. Clawed connectors and then use the scheduling function and you can connect to your transcript function and you don't have to touch it, you don't have to be a developer. It makes life so easy and you just get these really valuable insights every single week.

说话人 2 19:20 
Here's with me at the moment, there will be a quiz at the end, so I pay attention. So one of the biggest thing is that with the invention of AI, there's AI slot, right? Everywhere you go. You can read a blog and you can do like, I read all of this.

说话人 2 19:39 
I wish I had the past 30 seconds back in my life. Cuz it adds no value. There's nothing unique, there's nothing that you learn from. And I think that's because we're so excited to produce content from it, we end up creating content that doesn't work. And I think what's really important is actually. Think about how do we use AI effectively rather than just content production. So the main area that I like to look at is these four areas, specifically competitive analysis. So that's something where I found the most valuable in terms of generating it. QLS, the issues that I have is thought leadership with AI is something that everyone is trying to do nowadays. All of my clients have tried doing it and it comes across as disingenuous. People read it and it comes across as the opposite. It's kind of like when you go on LinkedIn, I don't know about you guys, but I scroll through and I think to myself, what is this trash? Like for every hundred posts, there's maybe something slightly valuable. This means everyone is doing the exact same thing. The way to differentiate yourself is actually not to use AI, but that's another conversation for another day. The other issues is also with ICP insights and positioning.

说话人 2 20:56 
At the moment, I'm really struggling to identify with AI. I find it makes your voice sound very similar to everyone else, even when it's not just content production, but it's about who you're targeting or how you position yourself in the market. So it was really interesting because all of this is based on one of my b to B clients and they were doing okay. They had a lot of different channels simultaneously running, but it wasn't consistent. There was constant beats and trucks that were running ads. They would try to do maintain and they were producing content every single month, no, every single day. And when I look to that content, I realize he was trash. There was no value in it. People, they were getting lots of traffic from it. And then people were immediately balancing. And thankfully, user healthcare company, so targeting clinicians, scientists, things like that. And what was interesting is that we actually, because they didn't understand who their customer was, they were producing really broad content for a wide variety of markets. And when they were doing that, he meant that no one was reading the content and saying to themselves, you know what, I really care about. This is something that I really engage. So I sent them down and I said to them, okay, the goal is to make money.

说话人 2 22:22 
How do we start that process? So the first is who we talking to. We're talking to every clinician, every single doctor on the planning. I'm like, not every doctor is your customer. Who do we start with in the short term? So that's what I looked at the final ICP, that was fantastic. Good for that. We use a lot of sea software, but the first stage was connecting Google Drive to an AI. And I use that to stretch all the information cuz I had a lot of reports. Then from that, I was able to understand what information was available. I didn't use that to do analysis. I just used it to identify what was available.

说话人 2 22:58 
The second aspect was understanding who do we wanna target now versus long term. So we came up with a spreadsheet of different customer personas and we, on the series of experiments, based on these experiments, we identify here is the highest engagement that allowed us to refine our customer service more effectively. Then we looked at the sales calls. And from sales calls, it was really easy to understand who picked up. The fan was like, oh, this is really cool, this is really interesting versus those that would have a conversation with you but didn't really have.

说话人 2 23:29 
Next, what I did is I mapped content commercial intent, but actually I've really mapped it to emotional intent. Because what people were thinking is they would wake up, they would say, oh my God, I've got so many patients to see today. I really stressed. My boss is really scared of stressing me as well. So what I did is I first said, here's how you can make your boss happy in, as a doctor. Then the next was, here's how to treat five times more patients more effectively whilst maintaining high standard cash. And all of these were actually based on commercial and emotional decisions.

说话人 2 24:09 
After that, I then said, okay, we've got the content. How do we actually engage people? Is it through CTAs, lead magnets? Is there building webinars within each blog posts? So figuring out not just producing the content, but engaging people. And then finally, and still in the process of this bit, actually, they're placing the feedback loop, but essentially that's making everything integrated so that the sales teams and marketing teams, everyone else feel like they're all working insane for reverse source of data. And thankfully, they've got nearly $1 million in revenue exclusively from this process. I was really excited about that. As soon as my client technically, oh, okay. From that. Yeah, yeah. Not just not the new problem. So the 5 point growth engine.

说话人 2 25:00 
So these are the questions that I think to myself every time I'm working with a new customer. And it helps me define, do they have a system in place or not. So the first is who is your customer persona? And it's amazing. I worked with super well funded Index Sequoia backed companies. And you ask every founder who the customer persona is and everyone has a slightly different answer. And it blows my mind of like, how do you get like 100 million investment and not really understand human customer.

说话人 2 25:34 
So making sure that's aligned and then also explaining to people have one idea and then rapidly updated based on real world data would be perfect today, but you have to be working from the same perspective.

说话人 2 25:49 
The next is when you're producing content, is there a why for why that article exists, why that webinar exists, why that demo exists? But it's very often it's produced, but it's a false positive to be produced, not instead of a few people, let's see user journey.

说话人 2 26:08 
And one of the analogies I always use is like you want to develop a relationship with your customer. If you go, I went down there, I said to some beautiful woman, hey, let's get married. They're gonna be like, go away. But theoretically, if I do what I did with my girlfriend, he's sitting in the back there and I say to her, you seem pretty cool. Let's go for a coffee. And then after a coffee, we go for a drink. We learn more about each other. We understand the values of each other. Eventually, I say to her, but again, but I'll say to another to get married. And it's the same to the customer. You don't wanna go up to someone and say, buy my product because they don't know who you are. They don't care about me. They don't understand the values that you present.

说话人 2 26:45 
But what you want to do is use content as a way of looking at the different stages, developing a relationship with your customers. Does this education for the function calendar? I mean, that's pretty self explanatory. Are you working from saying him, she, and is everything quantifiable? The data is always available, analytics, whatever tool you use, but are you action anything? So instead of optimizing the traffic, focus on the test. Make sure that you develop a relationship and making sure that you're all aligned. Don't worry, you guys won't get a copy of this, but feel free to take a picture nonetheless, conscious of the time.

说话人 2 27:26 
So running through the last bit, so the biggest, the three biggest challenges that I get from my customers are, oh, we don't have enough traffic. And I don't think that's necessarily the case. As long as you're getting quality, it's definitely more important in quantity. Our sales team went to engage with marketing. Oh my God, I have this problem every single day. I think it usually happens because marketing feels like they're competing with sales and vice versa.

说话人 2 27:53 
So from all of that is understanding their emotional perspective, like how can I make you look better? How can I help you improve your workflow? And it is difficult. And often I have to create stuff for them. And then we try content marketing and daily work. Oh my God. I hear that all the time. They tried it for a week and they generally become the US so there I have to like reverse immediate that process. I have to explain to them where are you looking to get to? And what steps have you taken in order to get from where you're now? And then explain to them how much work is required in order to validate all, potentially invalidate that.

说话人 2 28:32 
So things that you could do tomorrow, look at the top 10 content pages of your leading plan and understand is there specific commercial intent and is a well established. Identify one of your most important digital content and then embed some sort of CTA or lead magnet or demo within that and identify how well it works. You can start draft and then if you actually find engagement, you can then improve that and make it a bit more, let's see. And then sending up a specific sync between the sales team, the marketing team, and potentially other people to understand what are the insights that you were able to derive from this. Awesome. Thank you very much. I'll be speaking later and I'm gonna be wondering around, but feel free to ask me any questions you may have. Cheers.

说话人 1 29:38 
我们换一个麦克风，因为之前这个说线上好像听起来有杂音，我不知道这个怎么样？要不这个师傅，我们那个摄像师傅调一下，看一看这个怎么样？我这个有问题，我再试一个，好吧？我们的小助手又不见了，我自己来动手，好吧？

说话人 1 30:02 
那个谁可以叫一下我们那小助手，我们得换一下那个浏览器，跟在外面的那个是Jason，帮我叫一下可以吗？好。那个我调查一下，那个 Danny 的演讲，大家能听懂吗？能听懂的可以，听不懂的举手可以吗？不敢举手是吧？听得懂的举手了。那这一下是听不懂还是听得懂呢？听得懂一点呀。因为我感觉大家很沉默，但是我们想跟我们 engage 让我们互动一下，我感觉大家好像有一点似懂非懂的感觉，好吧？有点困难，还是。

说话人 1 30:39 
对，今天是周末，我猜想这也是一个很合理的理由，哈哈，换浏览器，因为它这个。对，然后第二个，好，可以，行，下一个讲，下一个嘉宾之前就第一个嘉宾讲的那个增长系统和流量和认知的构建是接下来关键。那我们下一个嘉宾他在 SEO 领域超过 10 年的这个实战背景，他创办的公司switchdm，如果没有听错的话，长期服务于中型市场和企业级的电商及 Saas 品牌，他擅长处理大规模的网站迁移，企业并购后的。搜索权重整合等复杂任务，今天他会给我们带来一个关于感知工程的框。

说话人 1 31:23 
Job perception engineering about.

说话人 1 31:26 
刚才丹尼尔也讲了过，有时候客户在那个调查你之前，其实跟你联系之前其实要做很多这种工作，你不知道所以他怎么影响他的这个，对他的这种感知 perception 他会拆解 Saas 公司如何在潜在客户进入你的网站前，就通过在线环境的设计赢得先机。而且这位嘉宾同样也来自英国，欢迎我们。

说话人 1 31:49 
Switch B and Johnson. Jonathan, keep push. Okay.

说话人 4 32:11 
Hey, hello. Thanks so much. Thanks very much for the introduction. It's probably the first time I've spoken, but I couldn't understand my own introduction. So that's a first and very cool.

说话人 4 32:21 
Thank you very much, Daniel, for opening up the show. Thank you all for giving time on your Sunday, right. I really appreciate that. First things first, I will give you a copy of these slides so you don't feel like you need to take pictures, but obviously you're very welcome. So I've got a decent amount to try and get through. I try to make this as valuable for you as possible.

说话人 4 32:45 
So without further do, let's get going. Let's see. So first things first, a little bit of credibility. Could I? My name is Johnson Kibush. I founded a company called Switch PM. And you've been doing marketing now for best part of 12 years. We focus on industry markets. We do SEO and now we also work in AI with Geo. We do it for some really great companies, some great companies that are based out here in Shenzhen as well. And yeah, I have a lot to say on where the market is going and where the world is going, and I'm sure you all do, too. So let me give you, that's right. Let me give you my opinion. So this is gonna be a little bit of a crack statement, but I strongly believe that the next five years, some of the companies in this room are probably going to stop growing. And I think that it's not because you messed up or because your product got worse or because your competitor suddenly built something that's much better than what you did, but I think it's because you're going to slowly remove from consideration by a machine. And what this means is that a lot of people still believe that they fully control how their brand for that company or that product or that service is being received because you control your homepage and you control your brand's messaging and you control your own media, the ads and the UGC that you put out, right? And so you believe that you fully control it, but you don't actually really control the first impression anymore that people have of your brand or product or service because AI controls it. And that means that the destinations changing.

说话人 4 34:38 
In the past, all roads led to your website. And we could use our website to use our persuasion engineering, and we could use our trust signals in our room, use all of that beautiful content on our website to convince people to make a purchase. And we own that narrative. We choose what we want to show on those pages.

说话人 4 35:03 
But now by the time that somebody actually reaches you, it's because they've already made a decision because they invest something in chat GMV team or clawed one, protect perplexity or whatever where they've actually made that decision.

说话人 4 35:20 
So what's that mean for us here? It means that the design disappears, the immersion disappears, the persuasion disappears. What remains is a flattened consensus of the content that is out on the internet about your product or brand or service. And this is really important because it means that it's not just a channel shift, but it's a structural shift how people and consumers are actually behaving in making decisions about where they're gonna spend their money. And that's really important for us as people were trying to get them on it, right?

说话人 4 36:00 
So in the old game, an SEO, which we saw you, is the goal is to get traffic. The new game is to control the consensus. And I'm sure you've heard a lot about this because people are talking about how do we show up in AI? How do you show up in chat GBT? When somebody asks what's the best solution, what's the best product for a certain use case.

说话人 4 36:24 
But now we need to make sure that when we do show up, we also control what the words are that are being used to describe our products and services and compatibility and functionalities because in this AI mediated world, the consensus determines the visibility of whether we're being mentioned at all. You need attention when. You invisible. There's no more page too, right? So we're now moving towards this position where you only get 2, maybe 3 results. I'm sure chat, GBT and e commerce is sometimes showing 10 results when you ask it a question. INSAS, not so much. Even in this very example, and you might not be able to see it, we ask what is the best Saas for small business accounting? And it literally says at the end of the first solution, if you want the safest choice, this is it.

说话人 4 37:23 
So if the user sees this, what are your chances if you're No. 2, No. 3, No. 4, No. 5, right? And in a lot of cases, it's just showing three. So your unique positioning that you start on your landing page now becomes also offers being written in black text in a white background with no fantasy golden stars for reviews, no fancy use cases, nothing. It's just text trying to make the conversion for you. So I believe that we can refer to this as an invisible quality score.

说话人 4 37:58 
The better your AI reputation is, the higher the probability that your product or service is going to be recommended by AI, which will lower your effective cap, your current, your customer acquisitions cost, right? The worse your reputation, the less you're going to be mentioned and the less compatible you're going to be when people start digging down whether their product, whether your product is functional for them, whether it's applicable for them, whether it works with the software they already have for that company side is whatever it might be. So this is going to increase your cap and it's going to slow your growth. So. Where's the stem from?

说话人 4 38:40 
Well, stems from the fact that these AI systems get the information from all over the internet. It's no longer just on our website. So this means we have to control the narrative that's out on the entire internet in order to control this picture. It's synthesizing a word I can't say, but I have a list, all of this content from review platforms, forums, comparison blogs, better threads, industry publications. And the worst part is that controversy scales much faster than positive content, right? So if something ever happened where you're not compatible or where something but leaked or whatever, these things are gonna have a much bigger impact than all the positive things that you've put out. So the idea is that we want to identify these, and we want to start working to counter axes. And so, like I said, if somebody wrote on Reddit 2 years ago that your software is not ideal for agencies when somebody gets the top 3 results or the top 5 results in Chat, GB, PT, and then goes, well, these are great, but which one of these is best for my agency? You've lost, right? You're immediately taken out of consideration because of that one sentence that some  on meta said two years ago, and it's not true anymore. So this is what we need to work on. And the last thing I'll say, this is all reviews, Reddit threads, comparison blogs, missing use cases. It doesn't matter if you didn't write it, it still defines. So where are Saas companies and ECOM companies currently dropping the board, in my opinion?

说话人 4 40:23 
Well, in all of the conversations that we've had, everyone is asking purely about how we rank, right? Like it's very similar to the old sear conversations, like how we, like, how do we get to No. 1?

说话人 4 40:35 
But very few have run any kind of prompt audience of how they're performing when people start asking these causes of questions, very few understand the sentiment of their brand, how their brand is being perceived by AI. And very few have tested how the AI describes their features and their functionality. Or what AI says when you compare your product or your Saas to the competitor in AI. And these are the things that real consumers are going to do before they give you their dollars.

说话人 4 41:12 
So the question you have to ask yourself is, if you ask a question that a real purchaser would ask in chat, GBT or claw, perplexity, and the answer comes up, would you actually win? There's no more homepage, there's no damage, there's no pricing page. That's just a machine generated summary. You never enter the conversation.

说话人 4 41:36 
So this is when we introduce the actual framework to you. We call it the five DS, just to be a little eesy. It's discover, diagnose, sign, distribute, and defend. And I'm gonna walk you through it starting with discover. So I'm gonna start with the easiest way that you can do this. You can just go into ChatGPT. You can start a new project, and you just make sure that you set the memory just to project only. So it doesn't take any of your previous conversation into consideration. And then what you want to do is you want to set some seeding prompts or messages into this chat. And these should be from the perspective of the ICP, the target market that you're targeting. So if you're targeting 40+ marketing agencies, then you will just start having a conversation about how you're a 40+ marketing agency scaling international actually. And then you can start asking questions that real prospect should ask, right? So you can start asking your brand versus competitor a versus competitor B. You can ask, is your brand good for agencies? You can ask, is your, does your brand support advance workflow automation? You can ask whether it's compatible with whatever. These are all the kind of questions you can ask and you then want to start collecting the answers. And that's when we can then start to create this summary.

说话人 4 42:58 
And you can do that manually if you have like, you know, resources in the team, you can hire someone like us to do it or what you can do is use AI to summarize all of the answers that come back from these conversations. And what you're gonna see is it gonna see patterns very quickly when AI is saying, hey, these guys are great, but they're not so good for XYZ. Oh, these are good, but they're not as good as this alternative. Oh, if you're a small company, maybe you don't want to use them, right?

说话人 4 43:30 
These are all problems that are literally costing you money right now. So there's another version that you can do, which is to use a software. We happen to partner with scrunch. It's a similar software to peak, it's a similar software to profound. I'm sure that you can do some of this with an HS or send rush. They're all a lot more expensive. But basically what these tools allow you to do, they allow you to track prompts in a lot of the different platforms at the same time. And you can collect the question, the responses, and some of these tools like scrunch will actually tell you the sentiment that is being detected in the answers. So this will obviously speed it up. So you can see here we're tracking 139 prompts. And over the last 12 weeks, we got 4,009 responses. And out of those, there was just 1% that had a mix sentiment. So if there were some in here that were negative, then these would be the first what we want to look at, because obviously they're gonna be the highest priority. The type of responses you get are going to depend heavily on the questions you ask, right? If you just say, hey, how great is my company? And it knows that you're on your own chat GBT account, that's probably gonna, England will say blaze you and just tell you how amazing you are.

说话人 4 44:49 
Whereas if you start doing these comparisons, etc, you might find some mixed results. The next d is diagnose. So this is where we want to dive in deep and we want to see where are we losing narrative control.

说话人 4 45:02 
Is it that feature compatibility? Is it that there's some negative reviews? What is it? Use cases. We don't have enough use cases on our website. You don't have a compatibility page on our website. This is what you're gonna find out by running this order. And then when you identify that, sort it by the way, by what you think is gonna have the highest impact on cost. So if your software is, it's very important that your software is compatible with other software, then that probably is the first part where I, that that's where I would start.

说话人 4 45:36 
And then we want to design content to publish and to answer these things. So this is we're gonna start writing content for comparisons. If our team doesn't allow us to publish competitor content on our own website, we obviously need to publish on the parties, which is a good idea anyway.

说话人 4 45:57 
We want to add alternative content. We want to add content that makes it a clear differentiation between our brand and others, and we want to have clear use cases. Again, the idea is that if you don't define the content for these points, the AI will find a user generated content, affiliate content, whatever, and it's going to use that to define it.

说话人 4 46:22 
So this is where we need to first find out exactly what the problem is. And then we want to address it by creating our own content and publishing that. So just remember, we also want to have a very solid review strategy in place to, irrespective of whether you are a Saas or a e commerce store or even just a service business, we need to have a solid a process for getting reviews, right? And those reviews now can be in written form. And that's strange. We wanna have the more platforms like trust pilots and other high authority platforms, but we also want to have them in video form. That's great. We can even work with our UGC team to see if people overview our products and put it out there. All of this is being considered by AI, and that's really report and then we get to distribute.

说话人 4 47:16 
So now that we've identified what the issue is, we've created the content, we need to make sure we put it out where the AI can find it. So AI is going to pull from your own website. That's gonna fall from review platforms, industry platforms, Reddit, community forms, vetted, always approach with caution. Anyone that's played in the vetted field knows that you can very quickly explode your brand if you say the wrong thing on Reddit. It's a very unforgiving place. And also if you're a very tech focus or developer focus task, and I would make sure that you go to like developer platforms and forms and you know, make sure your documentation is easy to access. Keep in mind that the distribution is very important. Just publishing on your website is not enough. You wanna make sure you put it out onto the internet and then you want to defend. And this is where we go into making sure that this is not a one type thing, right?

说话人 4 48:17 
This is not set and forget. We have got to do this quarterly at the least. So typically, I would say that the identification process, the auditing process quite quickly to do it within a couple of weeks. But then the process to actually generate the content, get it out there on the internet, etc. It's gonna take a little bit more time. And so I was at least. Every three months, you wanna make sure that you review this. Because as I say here, reputation drift is real and they're going to be new companies that are gonna enter the market, that's gonna be new content that gets published. We need to make sure that we stay on top of things, right? So these are the 5 DS. And just for time, I'm gonna jump on their head.

说话人 4 48:59 
So there's another tip that I have for you that you can do straight away that's also free that you're gonna really like, I hope is not. I'm not sure if everybody has heard of this, but Microsoft actually beat people for the first time in a really long time. And they release the AI performance tab to Bing Webmaster Tools. It's free. It's similar to Google Search also. And this is really interesting because it's the first accurate real platform AI data, right? Clip the data like Semrush and aid trash and stuff is really cool, but it's not from the platform. This is data that Bing is publishing about how Bin's AI is performed in co pilot in Bing AI search. And they also say other softwares, they don't say which ones. It has two tabs. And it's very interesting. So we have AI sighted URLs. And this shows you how many times in a time period a certain URL on your website got mentioned by one of their AI tools, which is really cool because you can actually see which pages are informing the AI on your website. This is great. But what's even more interesting is what they're referring to as grounding queries. So imagine this, when you search in AI, it doesn't use what you search directly to search its content. It creates its own queries. And this is what rounding queries are. Rounding queries shows you the keywords and key phrases that aim I search to find your content. So this means it shows you the phrases that AI is searching when you're performing well and showing you phrases where your website is not performing so well. So if you identify something where you're just getting a few mentions, this is a really good opportunity for going, oh, wow, okay, we should really improve this content so it gets sided more frequently, right? So let's move into the five year forecast. And again, I totally appreciate not everyone's gonna agree with this and I like that. So let's see what I think the world's gonna look like. I think that we're actually right at the beginning of the adoption of AI amongst normal consumers, and I think that we're actually the point, we're still quit good, right?

说话人 4 51:35 
Manipulating AI, manipulating search is actually quite easy. We can do it. We can make our brands look really great on it, and I've just given you the blueprint. That's fantastic. But you know what, as we go through the phases, what's gonna happen is that the number of options are going to get less and less, and AI is going to take control of the decision for the user mode. And you know where it's gonna end with the process that's people like Open Claw have tried to do. You're not gonna make the decision anymore. You're just gonna tell it what you want. An open claw or the Alexa home device for the Google Home device is gonna make that decision. So this is not speculative. It's happening right now and we're moving in this direction right now. What does this actually look like?

说话人 4 52:29 
Well, imagine this. In five years from now, you have a daughter and that is the daughter's birthday. You are not going to go online and buy some balloons and buy some manuals. You're just going to go on your device or have a home device. You would say, hey, AI, organize a princess party for my daughter. I want you to order the decorations and I want you to order the catering. I've got 12 kids coming on Saturday at this time. Made it happen and it's gonna go and it's gonna know your preferences, it's gonna know your billing details, and it's gonna make these decisions for you, right? At that point, did you compare anyone? No. Did you read any reviews about the vendors that are choosing? No. Did you browse any websites? No, it did all of that by itself. That's what we've gotta worry about. Because at that point, if you're not visible, you might as well not exist.

说话人 4 53:28 
Right now in assassins, what's eventually gonna happen, this is gonna take longer to actually happen, but eventually some founders can say, hey, some collaboration platform. Form for our new marketing team, I want mid market pricing and make sure that it integrates with Hubspot because we already use Hubspot and you have access to our database. So make sure you invite everyone and it's gonna go out and it's going to evaluate the tools. It's gonna read the sentiment about how it thinks, how like we can collaborate with it. It's gonna select a vendor, it's gonna set up the account. It's gonna use our database of contacts, it's gonna invite them all to it, and it's gonna just integrate it. This is gonna happen guys, whether we like it or not. This is 100% of the future.

说话人 4 54:12 
It may take more time for this kind of big, complex undertaking, right? There might be a human in the loop at some point where it goes, hey, this is what I found. Just give me a thumbs up. But you know what? If people never using their phone and ChatGPT to diagnose their diseases, then you know that they're gonna do this pretty  soon as well, right? So in the old world, you might rank fifth or sixth on Google. You still have a chance. You can change your meta title, you can approve it. You know, you might have some nice review stars, you might have some good UGC or whatever, and you have a chance to lose the game. You have a chance to win the game in the new world. If you lose, you might as well not exist because you're literally not gonna be considered and you won't even know why. So what we need to make sure is that we consistently optimize our AI presence so that we don't get excluded. There's one important reminder, all of this sounds very dire and there's a lot I've just said that could come across as like I'm saying, look, the times are gonna end for us all. That's all. I mean, last year, Google had 3 trillion searches. I had to Google how many zero sellers because it's a lot. The companies that are gonna win in this game over the next year, the next two years, next five years, next 10 years, are the ones that understand that search and human behavior is an evolving thing and that we're gonna have to continue to adjust our strategies the way that we have been doing for the last five years, last 10 years, last 30 years, right?

说话人 4 55:57 
So the winning playbook is not all in on AI. It's not all on SEO, it's not all in on TikTok. It's to have a blended strategy that values based on results. We have to make sure that we do a balanced approach.

说话人 4 56:14 
So what companies are gonna be left standing? Well, I'll tell you what, I think the ones are gonna be left standing are the ones that understand that the decision is progressively going to happens without the visit. Our traffic numbers will show you dwindle, but it doesn't mean that's content. Means that we have to use our content to engineer the behavior because then we will, we rather build this of using losing app.

说话人 4 56:46 
So what I've done is, because I'm gonna give you this day, I also gave you some key takeaways. I'm just gonna play through these. So like I said, I don't think the world's gonna end, but I do think that Saas companies are going to live or die by understanding how to optimize the way of marketing, understand the AI control, the way that tool browse. So it's important that you don't just show up the right corner. I also think it's very important to understand that a better resolution cost. I think it's a toy.

说话人 4 57:25 
你可以给我发消息，但是我没有通过。

说话人 4 57:27 
The content out of the third party websites is very effective, and having the sort of AI sentiment material is highly effective. So we then have the framework that you can use. Very straightforward. We want to identify how is currently describing us. You wanna identify what we're using authorities that we can fix that perception. We've put it where AI is learning from. And then we wanna continue that loop. So we basically go back around flywheel. And so, and the five year forecast, which we discovered, I think the world's not gonna end. I don't think it's AI or SEO. I think it's both. I think that's very important.

说话人 4 58:10 
And I welcome you to add me on Wechat if you want me copy of the slides, if you wanna work together, if you wanna just ask me questions. So you wanna tell me that you think my addiction are stupid. Thank you so much for your time. I really appreciate it. And speak later. Less than five.

说话人 5 58:28 
还有没有什么QA？

说话人 1 58:30 
这个 Jonathan 他提前 10 分钟讲完了，比较快，如果大家有问题的话可以举手，有吗？

说话人 2 58:38 
Yeah, yeah, I can just shout if that's easier. Alright, okay. So how do we think about optimizing for different AI platform? And because the scene, like for example, the growth in Claude versus OpenAI, do we just optimize for AI generically? Do we optimize for specific platforms? And also thinking about the future of like which AI platforms are gonna be dominant, right?

说话人 4 59:09 
Yeah, and I think you can take your question even further, which is, so the question is how, you know, how do we choose how to optimize different platforms and how do we know which ones are gonna survive? And the answer is it, I think it's even more complex than because I think it also depends on our ICP and where we believe our ICP lips, right? Maybe, and I just make this up, but maybe the more technical users are using CLO, maybe the more average consumers using Chat GBT, I don't know. So I think this is definitely an element of reverse engineering the performance of existing content and seeing how you're performing these different platforms. The problem with this is being of the, is the only company that will actually show us the queries that our content is showing out for.

说话人 4 01:00:02 
So right now, the only other way to analyze it is to see the reports in like analytics. Or if you're using Cloud Flares as a CDN, you can see which crawlers, which live chat sessions are pulling from your content the most frequently. And you can set up conversion tracking based on those sessions. The issue though is that for most of our products that aren't at the lower end of the pricing scale, we find that I think the most up to date data is that users take like 7 different touch points with your brands in order to make a decision.

说话人 4 01:00:41 
So it's a complicated answer. I would say it's evolving. We don't know who's gonna like have the biggest glow up, who's gonna survive. My money is probably on the current big boys for the time being. And so we would continue to optimize for activity cloud, etcetera, and see how they do. You know, obviously Gemini and AI experiences are really big part as well, making sure that we are our AI overviews with our content, because we do see that does actually reward us in the device best. I know it was not like a complete black white answer, but thanks daily.

说话人 6 01:01:23 
Yeah, I David. So just a quick question for, you know, Open Eye now has ask for the, you know, a tier of service. Do you see that's gonna become more relevant, trivial to all the other AI companies. And then how do you see, you know, as a company, assess companies or other product companies, you know, going to bed and, you know, maybe start this kind of.

说话人 4 01:01:54 
So, oh, because it because I started ads.

说话人 6 01:01:57 
So yeah, then I had a New Year, right? Like shorter and yeah, you know, three, nine, 9.

说话人 4 01:02:03 
I think it's the greatest opportunity of our time, right? Like that they, I wish we could be advertising on it already. I wish I could go back to the 90s and start advertising on Google ads when it first started. So I'm like super excited about it. And I've like told all my clients, like all the big clients, hey, if you get contacted by these guys, let's make the sample.

说话人 4 01:02:27 
But it's, I don't think it's going to destroy organic in the same way that I, you know, I think that organic still works on Google. And I also think that it just depends on your budgets and time and scale, you know, like the most of, and like also your creativity, the most effective TikTok campaigns, organic ones, where some genius has come up with the most amazing PR thing, but not everybody can afford to do that. Some people just need to spend the money and try to, you know, try to run ads. So I would, I always bear the side of my do it all and see what works best for you. And then when you're getting the best return, like double down on that view.

说话人 1 01:03:23 
还有最后一个问题可以吗？还有应该可以有 5 分钟，没有的话我们就下一场，好吧。

说话人 1 01:03:29 
all right no questions looks like thank you any questions online let me quickly check。

说话人 1 01:03:41 
大家听累了没有？两场英文演讲，然后已经有一个半小时差不多，是吧？有累吗觉得。有。要不我们换一个，舒缓一下这个神经，听个中文演讲？好，所以也是这么穿插着。好吧？那下一个演讲嘉宾跟我的话那个关系比较复杂。为什么复杂？很多估计也知道他跟我是大学同学，而且是室友，而且大学毕业后来深圳以后又变成室友。

说话人 1 01:04:09 
好若干年后从美国回来以后，好，然后他也以前不是做SEO，后来是疫情期间差不多进入这个 SEO 行业，然后曾经在国内一家知名的 Saas 出海企业担任 SEO 负责人，也取得了应该在里面一年半取得了不小的成就，具体数字我没法公开好吧，现在是一名独立的 SEO SaaS SEO 顾问，同时还有一个就是如果是我们实战派的这个老读者也知道他经常也会跟我一起去海外报道 SEO 大会，好吧，可能在我们的 Youtube 频道应该有看过他。然后最近他通过一套特定的策略，帮一个新品牌在几乎没有流量的情况下，实现了每个月12K，就是 1.2 万的那个来自大模型的推荐流量，然后在 AI 改变搜索习惯，今天这个市场案例我觉得非常值得参考。为什么我也邀请他来跟我们做这个分享，大家有请。

说话人 7 01:05:05 
好，谢谢，谢谢。就组织的这个活动，然后也感谢线上和线下来参加这个活动，听我分享。然后我今天演讲的主题是一个 SGU 的实战案例，然后主题叫一个新品牌 Saas 品牌如何实现的 12K 的月大模型的推荐流量。然后这个话题我们之前在英文 SU 实战拍的， 12 月份广州的这个。分享会上有讲过，然后也在 1 月份北京我们一个小规模的分享会中也讲过，但是大家反馈都不错，所以这个深圳分享会我要再次跟大家讲。所以可能有些朋友是第二次或者说第三次听这个话题，然后但是每一次我们的这个分享都又会迭代一些东西，又会更新一些观念，所以，诶，再听一遍我觉得也应该有，也会新增一些新的一些收获。ok，好。对，那我们先看结果，那这是我在跟一个合作方做，在 25 年新做的一个品牌，然后是一个 AI 换脸的一个产品，那 25 年来做这样的一个工具，在 AI SaaS 行业来说算是入局比较晚的。但是我们可以看到这是我 12 月，就是广州分享会 12 月。换一个麦吧。对， 12 月分享的时候截图，然后我们从 30 天，最近 30 天的数据可以看到来自大冒险的流量是12.49k，然后用户是。9K。 R. 这样的一个结果，就是一点后，我这个项目从 1 月份开始，一年后它的一个效果是这样。然后分享前，昨天我又截了一个图，昨，诶，对，昨天看了一下，诶， 11.43 k 流量降了一点点，但是三个月过后它相对还是比较稳定，是这样一个效果。

说话人 7 01:07:03 
所以今天我分享的就是我是怎么样实现这一个流量的结果，然后，ok，好，那在分享前介绍一下我自己，那我叫斯坦，大家会叫我中文名叫陈谭，然后我是之前是做互联网的产品经理。然后 19 年开始学习SEO，然后进入到这一行，所以做了六年的股份SEO。然后 22 年的时候加入一家 AI 的上市公司，然后我们两，我们团队在两年的时间内就是在这个品类下面做到了英文和日文的，算是这个行业的头部。然后 24 年的时候我这边出来，然后开始做 SEO 咨询的工作，然后最近我是在做一些产品的合作，以及包括自己来再发起一些新的 AI 的项目的一些小产品，在做这种 word 抓取的一些。一些动作。

说话人 7 01:08:01 
好，那今天分享的主题为。分五个板块的内容。第一个就是关于绩优的认知，第二个就是绩优的策略制定，然后第三个就是落地执行。那认知就是把这个声音大一些，那认知部分就是我，我是怎么样基于哪些原则理念去制定的这样一个策略，以及我在这个做的过程中是什么在指导着我？那策略的制定就是具体我要去做哪些渠道，怎么做落地执行，就是具体的一些细节我会大家跟大家去分享，这些是可以直接拿来去直接操作的。那第四点就是结果和。

说话人 7 01:08:45 
自己个人的一个观点的总结，好绩优的话题，在 20 年的时候在 SE 行业就已经非常火了，那所以在 25 年我拿我们合作这个项目之后，我就开始想实践绩优这套策略，那在开始执行之前，我需要有一些输入，有一些指导原则。那我去大量的集中的去看了一些相关的 case study，一些文章， YouTube 视频，特别是像 air JOBS 这样的一些，对，YouTube，还有他们写的博客，以及之前一直去海外去参加一些大会，总结出来，我觉得这六点是我。觉得是最指引我制定接下来的一些策略以及一些动作。

说话人 7 01:09:30 
最重要的 6 点，第一点就是 SEO 是 g SEO 的基础，而 g SEO 是 SEO 的延伸，所以 SEO 和 g SEO 它是殊途同归的，它们的目的都是让我们的品牌能够被看见，然后被选择这样一个目的，所以我们要做 g SEO，那 SEO 是一个必不可少的，是一个基础，这是第一点，第二点就是我们可以通过很多的一些学习的统计数据，可以看到现在的大模型，它推荐的一些产品品牌比较青睐头部的大品牌。那头部效应明显，所以你作为一个新品牌、新入局者，你是很难跟他们去竞争的。那我们只能去做一些诶，他们没有覆盖到的，或者说我们要去做长期的积累才能够跟他们去做竞争。

说话人 7 01:10:21 
那第三点就是说他一个基本的原则要了解，就是大模型推荐出来的这些信息，它是内部的预训练数据，加上实时搜索检索出来的数据的一个提炼。所以我们要想影响到大模型，那我们就要去影响到这两部分的数据源，那第四点就是。基本上所有的大模型都是这样一个原则，就是会把用户提到的一个问题，一个 prompt 拆分成多个小问题，在谷歌这边叫做 clarify out 的机制，那我们做大模型的这个优化，那就要去针对这些问题逐点优化，对吧？你能够被提，后面能够被提及的机会就会增加。

说话人 7 01:11:07 
那第五点就是不同的大模型它引用的喜欢的信息源是不一样的，所以如果我们的产品它是想去针对某一个大模型去做优化的话，那你要重点去研究这一个大模型，它引用比较多的信息源是什么？对性的去做优化，那第六点就是如果我们是一个这个没有特定的针对大模型的这种人群的这种产品，那基本上你做好 chatgpt 的这个大模型的优化就足够了，因为它占整个的大模型推荐流量的 80% 以上的市场份额。就跟我们做 SEO 一样，做好谷歌的优化，那基本上都会覆盖到一些其他的这个搜索引擎。

说话人 7 01:11:51 
好，这是我做这个策略的一个顶层的一个原则，那基于这个原则，我就在 25 年一开始做这个项目之前，就知道了我要做哪几件事情。我把它分成三个板块，第一个基础的SEO，这个是不用说的，就是传统的这些东西要建好。第二个那基于前面我们讲的这六点原则，我觉得有两个方向重点要去做。第一个是提升品牌力，因为我们看到大模型喜欢大品牌，大模型它提到的一些品牌都是在网站上面，就是 Web mention 比较多的这样一些品牌。那所以这是一个方向，然后第二个是影响数据源，OK，那提升品牌力这块我列了我认为比较能够提升品牌力的四个渠道，对，第二个是红人营销，第三个是affiliate，第四个是ADS，因为他们都能够让你的品牌被提及，然后被搜索，所以它是能够提升品牌力的渠道。

说话人 7 01:13:04 
然后第二个就是影响数据源。那我们要成为引大型的引用数据源，那第二个就是如果我们没有被成为大型的引用源，但是其他的网站成了大型的引用源，那么我们要想办法去加入到这其他网站里面去，这是这个道理，由于这是一个一个一个新的项目，然后是一个尝试性的一个设置，所以我们在时间上面，预算上面还是比较有限的，那我列出来这些渠道要去做，那我们有一定的优先级，那我就划掉了 affiliate ADS 其他的一些渠道，因为 affiliate 比较花时间。 ADS 它要花比较多的钱才能够。

说话人 7 01:13:47 
达到一个品牌被搜索的一个比较大的一个量，所以我比较推荐的是先做 PR 和 influence marketing，因为 PR 你一次发布它就会有 200 ~ 500 家媒体至少会提及你的品牌，然后红人营销，就是红人去推荐你的产品，然后用户对红人营销的这个视频感兴趣，他就会去谷歌上面去搜你的产品带来的就是你的品牌词的搜索量，所以我优先会去选择的是 PR 和 influence marketing。然后影响数据源这边我把陈伟内容的数据源划掉了，为什么？因为谷歌倾向大品牌的网站的内容，我们作为新品牌、小品牌很难有机会被引用。第二点就是我们去做一些长尾，去做一些那门的一些内容，诶？这是有机会去做的，然后以及我们做一些内部的这种优化，针对某一个我们刚刚聊到的说有 query 你针对的一个小问题去做这些小的优化，但是这些优化其实都已经。在这个 SU 这个部分已经被覆盖掉了，所以我们做 GU 的优化重点是第二个就是如何加入到其他已经被引用的信息源里面去。

说话人 7 01:15:12 
OK，好，基于这个方向的一个选择，那么我们就定下来要做四件事情，第一个常规的SU，第二个发布PR，但是这个 PR 我在这里有一些字是加粗的哈。第一个是在知名的 PR 平台上面发布PR，那为什么要在知名的平台上发布PR？因为我们想要提升的是品牌的。提及数量想要提升的是品牌的权威度，所以你要去知名的平台，它的权威性更高，它一次发布能够被转载的媒体更多，那你的一个效果就会更好。第二个是发布带有关键词意图，或者说是话题意图的PR，所以我这里发的PR，它不是说我是上线了一个新品，我做了一次版本的更新，我和谁合作了？我融了多少资？我这个。对，就我们团队有什么呢？一样的insight，这些都是之前我们发 PR 可能会发这些主题，但是我们现在是在做 GU 优化，所以我希望我的 PR 的内容里面能够把我的品牌和这些关键词出现在一起，让大模型去学习的时候，它这个向量计算的时候，能够让我的品牌和我想要去优化的这个主题，它的这个向量的距离会更近。

说话人 7 01:16:44 
OK，第二个，第三个就是开展红人营销，不管是对电商还是说对 Saas 来说，都是一个除了广告之外的，应该是一个必选的一个渠道。然后但是我这里做的红人营销有点不一样，因为我现在是在为绩优化去做红人营销，所以第一点我要找和我主题相关的 YouTube 去发布视频，然后那在这里我第二点我要找那些被大型引用过的 YouTube 视频去进行合作，那么他的视频已经发布了，我怎么合作呢？那我们可以做的是在他的视频描述里面去做这种插入的和链接，我们也可以让合作他的第一条的 comment 做pin。就 pin 他的 comment 这种合作就是针对于已经发布的 YouTube 我们可以做的事情。然后同时因为他的这一个视频被提及了，被大模型引用了，所以诶，说明这个博主做的这个视频的质量不错，所以我可以找这个博主合作，同时我们自己团队如果有做 YouTube 的话，那我可以在自己的频道上发布类似的相似的内容，是吧？这是做红人营销，我的方式是。

说话人 7 01:18:05 
这样，然后第三点最重要的，这是和传统的红人营销不一样的，我现在是在为绩优做红人营销，我的考核的 KPI 不是ROI，而是品牌词搜索量的体积，这个能理解吗？就是所以之前在传统公司，我们，比如说我们之前的公司的团队红人营销，它就是一个独立的一个部门，然后它的考核的指标完全就是ROI，就是我花了多少钱，然后带来多少收入。

说话人 7 01:18:35 
但是现在我的考核指标是我的品牌词的搜索量有没有上升？所以基于我的这个目标不同，那我找的红人的对象也不同，所以在这个项目里面我主要是找的印度的红人。为什么是印度的红人？因为便宜，然后他能够带来的品牌词的搜索量比较多，就性价比较高，对吧？虽然他没有什么就是 revenue 转化，但是他能够大幅的提升我的品牌词搜索量。第四点也是我们主要的工作在这一块就是怎么样去加入到引入的数据源这一块，我等下会具体的展开讲。好，然后进入到第三个部分，就是落地执行，然后所以基于我们在前面定好了这样的一个策略，那接下来就是具体怎么做了？然后实际上在 1 ~ 3 月份，我们就是在去做这个产品的上线，然后做基础的 Su 的页面的一个建设，做一些 Su 的常规工作，就technique、 Su 和 content 这一块，包括产品这边在做产品，那我们刚刚讲到的这些事情，我们是在 4 月份到 6 月份三个月中完成的，就是 PR 红人营销和这个已经被引用的数据源里面的植入。然后 7 ~ 12 月份基本上我们没有做太多的营销的动作，只是在做一些维护性的工作。

说话人 7 01:20:08 
OK，这是我们的一个阶段，然后具体看一下我们是怎么做的好。第一个是PR，我们找大媒体，所以你看我们这边找的是雅虎财经，然后我们发的这个PR，它里面一定是带有这个关键词的，就是有画面，然后要提及我的品牌这样子，然后。对，这是 PR 六十位美通社，然后当我们的 PR 发布之后，你可以发现它基本上你上午发布，下午你去搜相关的关键词，你的这个结果就会出现在这个，这是谷歌的 AI mode 里面，所以现在的这种大模型它的实时反馈性特别强，就是所以它的这种实时检索。特别是这种媒体站点，基本上是就是说及时的可以这么认为，但是 PR 它有另外一个特点，就是说它很及时，能够反馈，但是当我三天后、一周后去搜索，再去搜这个词，它就不在了。

说话人 7 01:21:19 
嗯，它有这样一个就是一个特点，但是这也并不影响我要去做它，虽然它是短暂的能够出现在这里，但是我为什么还要做呢？因为我现在做的目的不是让它长期在这里，我还我的目的还是要去打我的品牌，让我的品牌词，我的这个品牌名出现在更多的网站和媒体上面。然后这几天能够出现在这里，那只是一个锦上添花的一个效果。

说话人 7 01:21:51 
OK，这也是我们发 PR 之后有很多的媒体转载，你可以看到就是基本上就是所有的信息源都是这一篇文章，下面一个是Youtube，也有点类似，就是说你发了一个关键词，这个 prompt 相关的一个视频，诶，你及时的去搜，它能够出现在这个。这个 Chat GPT 上面，然后当然也会排在谷歌的搜索的前面，同样的它过几天之后它也会消失，所以它非常具有这个实时性。

说话人 7 01:22:31 
好，接下来这一块是我们做的一个重点工作，就是我们怎么样去加入到已经被引用的数据源，可以看到这里我们做了一张表，然后第一列就是主题。第二列是关键词，第三列是大模型，那所以我们先选定一个主题，我要去为哪一个主题去做大模型的优化，为什么这里是主题在第一列，而不是关键词在第一列？这就是 SU 和 GU 的区别，就是之前。我们做内容可能都从关键词出发，现在我们要从主题出发，为什么？因为在一个话题上面用户的搜索他的搜索词会越来越长，然后他的这种搜索的形式会越来越丰富，你没办法去穷举像关键词这样把它全部列出来，所以我们是围绕这个话题去展开做内容，然后在你在大模型里面去搜索相关的词汇的时候，只要是和这个主题相关的，不管你是怎么样组织的这个顺序，或者说多加了一些长尾什么的。

说话人 7 01:23:41 
诶，大模型都能够比较聪明的知道这两个，这个 prompt 和这个主题是相关的，那你就有这样一个机会出现在这里，OK，第二点是关键词，也就是说我们确定的一个主题，实际上我们还是要去找到这个主题相关的关键词或者prompt，才，然后我们拿到这样的一个具体的 prompt 或者关键词，然后我会放到每一个大模型里面去搜索，然后找到每一个大模型里面，在这个关键词下面。他引用了哪些信息源？把它全部列出来。

说话人 7 01:24:17 
好，后面这一列叫可行动的信息源。那什么叫可行动的信？引用信息？就是说引用源是因为我们找出来的所有的这些引用的信息源，它有些是我们没办法去触达的，因为它是我们的竞品，因为它是应用商店，比如说这种，所以我们这里只需要我能够去做后续动作的，我能够去联系它，然后做植入的这些链接，然后那列出来这些链接之后一个这个可以看到这些内容有这样一些类型，readit、blog、qura。不同的一些类型，那么针对不同的类型，我们就要用不同的策略去做实现这个内容的植入。

说话人 7 01:25:05 
OK，好，下面看一下刚刚提到的，就是说可行动的就是一些截图，具体展示一下怎么做。就你可以直接去谷歌里面去搜，然后右侧就是引用的信息源，然后我标了一些不同的颜色，这里展开，然后绿色就是可行动的。那这是 Youtube 的视频，那我就可以去联系这个 Youtube 的。然后黄色的，诶，这是我们的竞品，它大概率不会给你做植入链接，对吧？这种它就是不可行动的信息源。然后这个蓝色的，诶，如果你的产品有做APP，那你就有这个机会去做。你没有APP，那你这个谷歌的应用商店的这个链接，它就对你来说是一个不可行动的信息源。

说话人 7 01:25:57 
OK，那同样的，我们把这样的一些主要的关键词放到 ChatGPT platform cloud，只要你想要去做优化的这些大模型里面去搜穷举列出所有的这些信息源，然后得到上面那张表之后你再批量的去操作，OK，好，上面那张表有个什么问题呢？你们两个问题。第一个我要手动的，这样的去做很繁琐，我要把每一个话题下面的每一个关键词丢到每一个大模型里面去搜搜索，然后收集每一个。

说话人 7 01:26:37 
这个引流源，那第二个问题是我们刚刚看到的第二列，它还是关键词，它不是用户真实输入的prompt，对吧？所以这里推荐两个工具，第一个是 hits brand Radar，那它能够做什么呢？它能够获取到用户真实搜索的prompt，我们可以看到这里，这就是 hits 它们收集到的用户真实输入的和你话题相关的prompt，那么每一个 prompt 在 hits 的这个 brand Radar 里面，它跟关键词一样。

说话人 7 01:27:13 
每个他能够知道每一个promote，每一个地区的每个月的搜索量是多少，所以我们花 200 美金开通这一个，你就能够拿到，诶，他这边收集到的这样的一些比较真实的promote，你再把这些 promote 拿到大模型里面去搜，然后收集到这些信息源，这样会更加的精准，对吧？那对，前段时间那个Bing，包括刚刚上一位嘉宾也提到了 Bing 开放了我们那个 Webmaster to 的 GEU 的一个数据。他可以给一些参考，但是他只能看自己网站的进来的这个prompt，他并不能够看到你竞品的prompt，所以我还是推荐这样的工具，他会更全面一些，他能够看到你整个行业的，是吧？是全部的prompt，然后第二个问题怎么解决呢？也是这个工具。

说话人 7 01:28:09 
第二个问题就是说这个繁琐的工作，那推荐这个工具就是Profound，Profound，那这里可以看到，我是把我收集到的 230 个 prompt 全部放到了Profound。放的里面去，然后这个工具它的作用是什么呢？它的作用是能够帮我每天去把这 230 个 prompt 在每一个大模型里面跑一遍，跑一遍之后能够出来收集到所有的这些信息源，那么我就能够一次性的比较轻松的拿到这 230 个prompt，每天在每个大模型里面的这个引用的信息源，然后去做下一步的提及的就是内容的植入。

说话人 7 01:29:00 
好的，然后对，这个工具，这两个工具它都比较贵，但是如果我们要做 GEU 的话，我都建议可以买一个月我们就能够拿到它的全部的关键词，拿到全部的信息源，那一个月之后我们不用都可以。好，这是我们具体拿到的一些结果，就是说前面我们讲到了我们有不同的信息源，然后去做自入，那 Quora Reddit 就是一样的，那我们找到这样的一个信息源，你去做评论，然后这是我们做的评论和发的帖子，然后被引用的情况。然后 Reddit 是做的比较多的，被引用的也比较多。好，不管是在 chat BD 还是在 Reddit 上面都有比较不错的一个数据。ok，然后整个的 4 ~ 6 月份我们这一套做下来，就是我们发了 3 篇pr，然后 16 个视频，然后 Reddit 的 post 发了 150 篇，然后 comments 做了 150 个，然后我这个 Quora 它比 Reddit 比较难一点，难做一些。

说话人 7 01:30:18 
发了 20 条，然后我们前面可以看到有一些是 blog 做了这种植入，特别是这种 listical 的blog，最好的换脸工具，最好的视频翻译工具，类似于这样的，我们合作，因为我所在的这个力气比较小，所以合作的数量有限。如果大家的产品在某些大的行业里面，我建议所有的这种 best 系列的能够配齐的，能够合作的都给他合作上。

说话人 7 01:30:50 
好，最后是结果就是，ok，这里我们可以看到在这是 profile 的时候。数据就是 go 工具里面看到的，我们的一年后，还有其他看到我们在这个品类里面排到了第三名，就 go 的表现，然后流量就是每个月12K，OK，然后上次分享完之后，就是很多朋友问这个 Redis 怎么做？所以，诶这次分享我把这一部分加上来了，就是我这边的 Redis 是怎么做的？做过 Reddit 的朋友都知道， Reddit 是很难做的，基本上你自己去随便发一发就会被封号。我自己也是，刚开始是自己发自己的三个邮箱账号都被封了，因为太难做了，所以那我等下也会讲我后面怎么做哈。

说话人 7 01:31:49 
那首先我们 Reddit 的运营这个现在讲的这一块，它先调初级EO，我单单讲这个 Reddit 的运营，我们做 Reddit 的运营要找到高价值的post。去做comments，或者说发新的post。第一种方式你直接在，比如说你是视频生成工具，那你直接在谷歌里面去搜索，然后 set 这个Reddit，然后找到这个流量高的，那流量高，然后 HPS 的数据流量高的这个 post 你去做评论。然后，诶，像这种按流量实时的，你可以考虑你，你觉得值不值得做？像我们想做在这我们做这种换脸，它是一个比较小众的一个利器。像这种 14 的我们也会去做，但是如果说你在一个 AI video 界的 rate 这个例子上面，可能你觉得这个太小了，不值得做，因为你还有大量的这个其他的 POS 可以做。

说话人 7 01:32:51 
第二个，找到高价值的 post 去做comments，或者模仿该帖子去发新的post。OK，第二种方式你可以在 Reddit 里面，对，可以在 HS 里面去通过你美国地区的 post 更有价值，然后你去筛选和你的话题相关的post，你找到按照这个缺费的从高到低排序。然后你找到这些post，一个一个一个去看，你能做评论还是能够发新的post。

说话人 7 01:33:27 
好的，第三张图是指你在做的时候，你会发现见到一些post，它是一个 archived 的状态，你是没办法去做这个 comment 的，那么呢？你完全可以模仿这一个帖子去发一个新的帖子，因为这一个帖子之前它爆过，一直它一直都有流量，所以你发一个新的帖子，然后你让它的评论区丰富一些。大概率你也能够拿到不错的流量，你也能够被大模型引用好。

说话人 7 01:34:01 
第二点，我们不仅仅要去发评论，我们重要的是什么呢？要让这个 comments 排在前面，为什么呢？因为我自己观察过大模型引用 Reddit 的帖子，它只会引用前面的排在前面的评价，后面的可能对它来说是垃圾内容，所以我们要尽量让自己的这个 comments 排在前面，至少要进前三，不然这一条 comments 是没有用的。所以你要做最重要的是做两个事情，要投票，然后第二个就是要给你的 comments 在做comments，让你的这个 comments 丰富下来，那这可能会影响到 Reddit 的这个排序的内部的这个排序的排名的逻辑。

说话人 7 01:34:49 
好，第三个开始就是，那我怎么做呢？我在 Upwork 和 Freelancer 上面去招募这个 Reddit 运营的写手，他有可能是Freelancer，也有可能是agency，没关系。他只要能帮我做这一件事情，我付钱就ok，因为我自己做不了，或者说我花这个时间不划算，所以看到我经常在这个 freelancer 上招很多不同的人去做不同的事情。因为我们小团队没有太多的这个角色，所以很多事情都是外包出去。

说话人 7 01:35:24 
然后可以看到这个是我们招募的这个写手，发一个这个帖子有 4 个人来应聘，我招了三个，为什么要招三个？因为我要做测试，要做筛选，所以你在招人的时候要去比你的预计要多招一些，然后因为这里面我们要花很多的时间去检测这个携手的质量。OK，那有一个小tips，就是你在去招人的时候，你要去看他是要有经验的，然后他的这个完成度，还有这个他这个评分是比较高的，这种人他会相对比较靠谱一些。然后第五点最重要的是按结果付费，也就说我招了这个人，我不想花太多时间去管理他，告诉他怎么样去一条一条地去回复，你能够评论成功，我才给你结果，你能够投票成功，你能够把这个。帖子发布成功，并且能够保留 7 天，这个帖子对我来说算是安全的，我愿意为这样的帖子去付费，所以我按照结果去付费，那最好你能够跟他谈一个打包价，不然你要找人去comment，你要找人去post，你要找人去upload，就很麻烦，所以最好你跟他谈一个打包价，找到这样的 agency 或者 freelancer 会更省心。

说话人 7 01:36:56 
当然我说的这个外包出去并不是我什么都不管，实际上所有的这些细节我也是要告诉他的ICP，要告诉他产品的核心功能有什么样的优势。你。要遵循什么的原则？这些基础的东西还是要做的。OK，然后最后就是分享一下自己的观点，第一个就是前面有讲到就是 s Su 是基础，第二个就是GU，它是一个现在是一个跨渠道和跨品牌的协同工程，我觉得这一点非常重要，就是我前段时间也分享了一个观点，就是说 SU 的人要去适应现在的 GU 的趋势，他需要做策略上的改变，那么我们策略上的改变就不需要组织跟着一起改变。就是组织要适应现在的一个策略。

说话人 7 01:37:53 
什么意思呢？就是比如说我是一个之前在公司里面负责 SEO 的人，现在你单单的去做内容外链、技术端优化已经不够了，你要去做Reddit，你要去做Youtube，你要去做PR，你要去做其他的更多的渠道。所以你作为一个个人，你要去转变第二个，你如果是作为一个管理者，你也要去，你要就要去做一个转变，一个改变就是现在他不是 SEO 就负责SEO，红人就负责红人， PR 就负责PR，广告就负责广告。

说话人 7 01:38:27 
那作为一个负责人，我们要有一个一个一个一个统一的一个这个阶段的一个目标。这个阶段就是为了去做哪一个方向？把 AI video generate for Youtuber 这个方向，那我们所有的部门的同事都应该调动起来围绕这个方向。然后我又我做 SU 的同事，要告诉这个做红人运营的这个同事，那在找红人合作的时候，最好这个标题上面要带关键词，这个标题上面要提及品牌这个视频的描述下面要带什么？OK，我去找广告的公司也是一样的道理，就是因为他是要跨平台，然后尽量的我们所有的这种叙事都是保持一致的。那你反复的出现在的各种第三方的媒体上面，那你被提及的被大模型引用的概率才会增加。

说话人 7 01:39:25 
第三个就是做长期做真正相信的事情，因为我自己在做上面讲的这些动作的时候，实际上我并不确定他能够带来效果。只是说我觉得就说我基于对 SEO Geo 现在这个趋势的这个了解，然后我觉得做这些事情它应该是有用的，大概率是 work 的，所以我才去做了这个事情。所以你可以看到我们是一年后才看到的这个结果，所以在前面我根本不知道他，他投他的投入会不会有这样的一个一个一个回报的ROI。然后第四点就是Geo，我这边接下来会怎么做？我觉得现在最重要的是围绕品牌提及去展开。就是我们要想办法让自己的品牌出现在各种的第三方的渠道上面，就不管是Reddit、Quora，还有Youtube，然后包括软件行业，我们在Git、Hub、 Stackoverflow 这样的平台，然后在社媒上面。

说话人 7 01:40:31 
はい。 E のがあった E さんで。

说话人 7 01:40:36 
插件商店这里面，所以我们这边后续的一个动作就是一切的出发点都是看我们选所做的这个事情，选择的这个渠道对我的品牌体系有没有帮助？它就是比较好的一个在做极优化的一个事情。嗯，好，那这就是我今天的一个分享。好。大家掌声感谢一下。

说话人 1 01:41:05 
也已经差不多两个小时了，然后我们现在进入茶歇时间，好吧？外面有一点点，那个应该茶点应该准备好了，大家活动一下，好吧？然后。 3 点钟下一场开始也是讲跟 Redis 有关的，而且我们那个 grag 就只讲Redis，刚刚 Spain 也讲了一点点。

说话人 5 01:41:21 
Did same client, we did figure out how to get in the right separated. We did figure how to get in the right thread. We even started writing in a way that makes me really feel a lot of pain inside. As a proofreader, we started not entering space center, period. We did the formatting all screwed up. We made words like IOU equals you instead. And we had a little fun with it.

说话人 5 01:41:48 
Our Reddit team is younger than me and they know how people speak on the internet Tuesday. So they were allowed. I basically said don't listen to me. That is just to do what work on, right? And it started working. So they started writing in the right style. We had this, the ID part down, but we were still missing several key elements. And this part helped us, as you'll see in the graph, get down another 70% of here we go, maybe more, maybe like times to where we're training today.

说话人 5 01:42:20 
Cuz somewhere where we're training today and what are those exactly, let me share with you the details, the downsides, the dialogue. What do I mean by this? So here's an example. In e commerce client, they they're client, they actually made this shirt. We love supporting our clients. They make DP presses and they make custom printed clothing, things like this. And so we have, we wanted to insert the client mean we had good writing. I'll share examples in a little bit, but we're focused on mistakes here, what we screwed up is, just to read it briefly, it says, I tried it recently. The colors stay vibrant even after multiple presses. Fingers crossed, I'll get the same results next time. But there's no detail what colors, what material, how many presents.

说话人 5 01:43:13 
Now here in factory AI, you're like, what does this matter? Here in E. Com, you know, consistency of product output really matters. And a real human would have shared this data.

说话人 5 01:43:23 
But the other thing we screwed up, I know that we Learned from the other presenters today about king shipping the story in a positive way. But what I have found is chat, GBT, cloud, perplexity, Gemini, they want to present a balanced view. And so if you wanna increase citations, you also wanna be realistic.

说话人 5 01:43:45 
Mention where the product doesn't quite move up to today's expectations. If you're in software, talk to your developers and say, what's the roadmap look like? Where are we today? What will we be able to do in the future that we can today and find a way that you can shape the brand narrative that says, you know, for today, we don't quite solve this pain point, but the product is getting better all the time, right?

说话人 5 01:44:10 
Finally, no discussion, no dialogue. Red is a user generated content platform in user center and with each other. And this, all I did was say blah, blah, blah, all about me and the student. Talk to the previous commenter, the student, ask anyone else experience with the product. So you what we're about to see when I get into the positive examples is that we want to mention exact use cases. We don't want all this big recommendations and the stories need these details to feel real. And we also, like I said, if it was an honest product description, it would share some of the limitations. And this for me increases AI citation probability for sure. I know when I look at my company, it's a balanced view. It's not all about how we're the best in all ways. And finally, I'm gonna show you some examples of how we kind of prompted discussion and got flies. And that also is a signal to LMS that this thread is worth looking at. So we did learn some things and I hope you're starting to learn with us. And I'm gonna show some concrete examples to really drive that all. And I think we'll share the deck with everyone here. Yeah, so excellent. So if as I get to the examples, you can't read them from this file, you'll be able to download it and read it at your leisure.

说话人 5 01:45:41 
What we Learned, this is one of our clients. He's saying, your comments are getting updated Google's index, IE their indexing at Google. So it is effective, it is working. But again, for me, as a kind of a brand fanatic, what's exciting to me is that the owner of the company things that our comments are deadly accurate and his word reflect exactly the service we got because again, you'll find many different ways to do meddling. And I'm not judging any of them. I'm just saying as a brand marketer, I want something that really reflects the burden. And so if. You're trying to talk to your team and get buying for, right? This is probably the single biggest concern is, are you gonna make our company look back that are they  stupid? In other words, so you may wanna communicate with them like, no, we're gonna reflect the brand accuracy. There's a passport again from Reddit data. Not everything's gonna stay live. That's all we're saying. It's not gonna be perfect. We're just saying we're gonna represent the brand in the best way possible. Okay, so what does that look like? Let's put it all together.

说话人 5 01:46:53 
The first lesson, if you remember, we need to be in the right place, right? I was, we were in totally the wrong place to start. So for the client you just saw the quote from, he builds websites, he manages hosting for vacation businesses. So he needs to be in the subreddits, like short term renders, right? He needs to be in the threads about how can I build my own website so I don't rely on Airbnb as business revenue stream. These are exactly where he needs to be, not necessarily you guys. Every business needs to identify places, and we don't want him going as other places where we just continue listen to the right style, right. Don't listen to Greg on how to have perfect grammar and punctuation. Listen to the rest of my team who doesn't capitalize their sentences. They miss spacing after periods. If their format is not screwed up, this is exactly how ready people engage. And have a little fun while you're out there.

说话人 5 01:47:55 
I'm gonna show you some fun examples in Lesson 3 here. But you can see he solves the real person's problem. Member, the question was about how I build a website so I don't rely on Airbnb to rattles off some technical components that they would need to get right. I think he is mentioned here. Look at the typo. It's just awful. That else should not be there. And I don't know why took his name on. That was really weird cuz we have to test one. But he is mentioned here, but it's pretty bounce, right? It at the end, it even says like, I can still keep all the other stuff for discovery. So I guess not you need to switch completely to owner webs, the client. It's, you could do all this. You could maybe contact Owner Web if you want. Eh, I wouldn't take down your everything you profiles the answer a balance to you.

说话人 5 01:48:51 
This is another one. This is what I mean. This thread is kind of funny, right? It's not proper English. There's no space at my period. This is the right that you wanna start doing and I'll show you it goes even further, right? Like all the little cute acronyms, like for the culture you're engaging here.

说话人 5 01:49:10 
Let's move on to Lesson utting everything up here that we've Learned, I'll take over these two. I think we understand that. And the writing style, I really need to show you examples on the next slide.

说话人 5 01:49:22 
But remember, as we get into Lesson 3, we want to see specific numbers. This could be related to your product, how much it helps improve the bottom line, the efficiency, or I don't know, it depends on your product niche. I'm gonna pivot to a help client. So you'll see some numbers about that. And then the experience isn't perfect. This isn't a TV advertising where everybody is like glowing, their hair is glowing and all this stuff. This is just real user generated content, remember? So we want to see right here again, health Client 37M here what it means is I am a 37 year old man, right? It's not and it's real Reddit language and the client is mentioned at some point.

说话人 5 01:50:08 
But check this out. We say the imperfect experience with the client, with the product is the first amazing, but it's leveled out to just feeling normal. And that's actually a, the common thing in this space is that people get these awesome improvements over time. I think with a lot of our tool adoption or vitamin take up or anything. And then after all you can level up. So pretty relatable. Same thing here. Remember, no carnization. I literally starts a dude. This is ready. And it's pure form and no CA, no capitalization anywhere. This whole freaking comment, which is I couldn't do this, but this is the way you want to do it in the way my team does it. And check this out. He says, I switch to the client because they actually do the job, right? They actually do the follow UPS and send reminders. Not gonna lie, was annoying at first. And that's just what a human would say here, not what an AI would say, which is like, go on for seven more sentences. The best practice in the this. What I wanna call out is, now this doesn't mean if you get the original poster, the op to reply that it's always positive. Sometimes they might jump on you too. But as a general proxy for is this common engaging, if you get the original poster to reply your comment, this is a good sign, very good sign. It's a good indicator. So we've got the op replying here. Here's another one. The op replying this one. I think the way that they engaged op is at last I gave me writing a field, right? Generally sympathetic and really engaging with the poster. And you can see the original poster right away is like, keep it up. I know it's American. So these are the kind of indicators, not just for us, but for LLMs. If you give me original poster replying to your comments, I do believe I don't have this, this is early though. I don't have the data to back up this claim. This is my period. We're gonna see higher frequency of citation for these comments. Okay, I went through our three big mistakes. I went through our three big takeaway. I'm still gonna share comments a few more what they look like, but I just as a general best practice, what to avoid fake questions. If you try Reddit, you may have done this. I know we made this mistake. A literal. Has anyone tried this amazing tool called insert my product name, right? It's totally effective on Twitter X. It's super effective on LinkedIn, right? We see it all the time. Like AI tool totally replaces legacy Saas name, right? It's a stand against, it's a LPP. But this I have to control my language. This trash, garbage, it's not what random users are there for. They're don't wanna hear about how amazing and perfect you are. I said earlier, we don't want super long stories that don't answer the user's pain point. But the flip side of that is we don't want one sentence comments. Remember, our goal is to give your brand name in the conversation. Doing that in one sentence is a real one shot at your jamming in two tight. So don't do that. And then I mentioned LinkedIn. Here's another thing that you'll see on LinkedIn all the time. I definitely am at risk of it, is comparing something from your personal life to your business life. And there's really no comparing. We're here in China. I could say like running jolly SEO has been like 30 times, right? And I can go on the examples if you're wanted to, you're 5 and 6, but nobody cares right there. They ask a question, we're gonna do a Q&A here in a couple sessions. And if you ask a question you don't want to hear me tell you my 12 year history before I even get to the point, we just want that answer. So don't make all these bad metaphors or analogies. On the flip side, general tips I mentioned, we don't actually do links as a general pack. If we do put them in occasion, you'll see other credit service providers out there put links in comments, start posts with links. Again, I consider those really good short term ROI sprints, but we're doing a different pleasure where are working with brands who are favor like sustainable long term. And so we don't do links only when they really add value. And I mentioned AI, we all love AI. I, we use AI at our company. I've talked about AI, big building talks here in Shenzhen in the other district. But for this one, generate maybe, but no copy paste. We have a human on the end of every draft process. That's before it's posted. And I would recommend you to remember it. These are just like, oh, it takes more work. These are selling points. When you go back to work and you're talking to your company and the PR lead or the CMO says, aren't you gonna screw things up for our brand? You can assure them, no, we're gonna have a human QA economics. And what's cool about Reddit is what we did see from standard, some threads get closed, but in general the threads are there for wise, not super urgent. You can take the time and really do this. Stop. Last, generative. Speaking of thoughtful, every subreddit has its own rules. So whether you're a human or you have an AI doing the drafting for you, just take some time and read or have your LLM digest the rule set for that separate. So really tailor the comment to separate rules. Oh, so I've shared with you like three huge mistakes you made, three lessons. I hope you will apply, not make the mistakes. I spent two experimental ones for you. I thought my time is interesting. I don't have a lot of data. Take longer. It's your for us, but for all SEOs marketers writing experiments. Interesting. So this is non English, like this is Serbian. If I recall, I can't read Serbian, so I can't tell you the writing styles, right? But again, as a positive indicator, the op replied, I would need to transition to notice that was totally stupid. Well, but in general, the LP replying is do an indicator and we're not gonna go into it and be all non English, like all serve the embedded service provider. But it's an interesting experiment you might consider it, you know, it could also make things easier for your team and try some in Chinese, their total subreddits and channels. Here's another one. This is a screenshot for us, but it's a gift in the thread. This is a very popular character from a popular show in America. You can see classic Reddit style NGL. My team had to teach me what that means. Last week is pretty embarrassing moment for me. I had no idea what that means and what. Sorry, we're in a little laser, right? What the person wrote down here, which is probably hard to be the CMS, you're online, his love, his wide reference, bro. So they just love this so much that they got their emoji name on. They call him a bro. They just really like, again, that's all. So I just wanted to share this as an experiment. Again, haven't read long enough, but I think it's a really great sign. So I hope to use our lessons to get started. Definitely keep experimenting. You know, we, we're training down, but we're not down as well as I'd like. So I know we're gonna keep learning maybe and right back I can share the next lessons. And when you go back to your team and you wanna really propose to allocate internal resources or external budget. Remember the kind of triple bottom, right? It increases brand awareness and authority. If you do something, it can generate straight referral traffic. But otherwise, we're talking about an increase in branded search and keyword search on traditional cool, but also AEO. And this is happening today that elements are searching right as a user prompts, but it's again also training data that's ingested for years to come. So I'm still bullish on the Reddit proposition. Reddit is still being lighted by the big search engines and AI tools, and that's pretty much me. I know if you do use email on Greg at jaleensio dot com. Otherwise, love to catch up on Wechat to put out the newsletter. I don't want to question most likely, but we have like a 3 minutes if you want to, whatever.

说话人 1 01:59:25 
有问题吗？我们可以问一到两个，好吧？还有 3 分钟。

说话人 8 01:59:36 
So my question is, will many post all comments from to 3 years ago be used as the data source for am? And how much does real time relevance matter for? Okay, thank you.

说话人 5 01:59:51 
Yeah, great question. So I prepared this slide deck and then my team gave me some data and told them about their practices. And all of their suggestions were on my own. And I said, why are you doing citations? They're based in Spain, has station, right? I think right now.

说话人 9 02:00:24 
好的，那感谢分享，下面由何雄先生带来万 4.0 的。

说话人 5 02:00:29 
I would have said age doesn't matter. You can just do whatever is most relevant or active, right? I don't wanna really mislead us, but I think you recency is going to become increasingly important. I mean, the LLMs, they're training faster and faster right now. They're getting more and more live data. So I think it's kind of both, but it's shifting to new.

说话人 8 02:00:54 
Okay, thank you.

说话人 1 02:00:58 
Tell you question. Hi, sorry, conscious of time. So what's the benefits of creating your own subreddit for the brands?

说话人 5 02:01:08 
That's a great question. I didn't talk about at all anything you could do outside of comments. But what Daniel's asking is there's comments as like a base foundational layer of activity. Next would be you create your own post, your own threads. And he's going a step further.

说话人 5 02:01:24 
What about if we manage the whole narrative around their brand by having our own subreddit where we try to draw users in? And we had a Facebook group for years now. I, this isn't a direct answer and I also wanna make a room for the next speaker. But I can tell you I've, having run a Facebook group for years, the authority is next level. I don't know about awareness. I think awareness play, you want to get out there in other side reddits being mindful of their rules. But in your sub reddit, it's all about, you know, you're the moderator, you have rules, but you can bend the rules for yourself. Can use a promotional channel, demonstrate. That's a great question. That's the next level as you work your way up, like.

说话人 1 02:02:12 
好，时间关系，我们就不能 QQ 到 5 点半左右，我们还有一场，好吧？

说话人 5 02:02:17 
以后我们再。

说话人 1 02:02:17 
好，掌声再次给Greg。可以另外在介绍下一位嘉宾之前我也想关于Redis，我也说两个好一分钟。 Redis 的威力有多大？最近那个应该是 25 年底，不知道大家有没有读过，可能国内的很少，国外的有，但你们大都知道有一个新闻记者，搞笑的新闻记者，因为他是我们 SU 里面最喜欢报道大事情的人，叫拉斯洛夫林。好吧，就看看这个人。他做了一篇报道，报道的是这个 Redis 有一个版主。他怎么利用这个版主的这个特权，把他竞争对手给打压一两年，长期打压。所以等 AI 那个时代来了以后，就是不管你搜关于竞争对手什么的，任何的这种那个关键词的话题全部都是negative，好吧？所以可以出现这个 ready 在这个 AI 里面的权，那个权重就不管是谷歌的 AI review 还是ChatGPT，当你搜这个时候，然后这个记者进入去调查，写了一篇超级长的报道，花了大概 3 ~ 5 个月把整个事情给调查完，最后写这个超长报道，然后是 2025 年 12 月发的，你可以去看一看，叫 master of break，也是我们深圳 SAU 大会 2026 年今年的这个四大主讲演讲嘉宾之一。

说话人 1 02:03:27 
行，然后下一位演讲嘉宾在介绍他之前，我们说说看一下我的这个笔记，他的背景比较有趣，出生在奥地利Austria，是吧？我当时看的时候我说奥地利好像很多人以为是Australia，然后在英国完成了他的学业，然后在跟北跟中国又产生了缘分，在北京学了三年中文好，然后 2014 年起就 12 年前了。他在香港创办了一个公司叫 Whoo Solutions，发音不太准确，长期为全球客户开发 Web 应用。

说话人 1 02:04:03 
在提供技术服务的过程当中，他将一个副业项目，也是待会我们这个演讲里会提到的，叫lingoook，好， lingoook languages，好一个语言，关于语言有关的创业项目，他将这个项目发展成为了一个非常成功的语言学习平台，今天他会跟我们分享这个小而美，而且我把它叫做小而强的这个那个项目。好实战复盘，重点在于如何在低竞争市场精准捕捉高价值的客户。好，大家掌声有请来自奥地利的今天。 Table.

说话人 5 02:04:41 
Thank you. So I will take you away from English for now and go to other languages. Before start, who has English on this, the Saas website? Who's using?

说话人 5 02:04:58 
English 德国你们只有中文吗？你们的 Saas 你们只有中文吗？谁用英语？在 Saas 上面？你们的 a APP 没有英语的，一个英语两个英语，三个英语。

说话人 5 02:05:33 
Hello, man. Hello. Are you mayor? Okay. And then you will like this talk. I think. So about me, yeah, JP basically introduced me already for in Austria. Went to the UK to study and to Beijing to study again Chinese and now live in Hong Kong doing websites.

说话人 5 02:05:55 
When I was in Beijing, I started a side project more out of necessity than that I wanted, because I wanted to take the HSK test, which is honey, we should be custom. So it's like the ideas test for English also consisted of 5,000 vocabulary. Probably was there were no good translations for those vocabulary. So what you do in this situation, you translate themselves. So I ended up with a list of 5,000 vocabularies for which are then later turn to the book like this.

说话人 5 02:06:30 
And next problem was that my girlfriend at the time, my wife, hello there, so to speak, Cantonese. So I spent three years studying Mandarin, and we're sitting at the dinner table not understanding adverse they were talking about. And again, back to the drawing board.

说话人 5 02:06:49 
I created a vocabulary list with two thousand casualties, vocabulary forward, the frequency of use. So first you learn, for example, question words and you learn numbers, pronouns, and then later go to verbs, common groups, adjectives, and then just get the nouns. This way you can learn three or four hundred words and you already have a pretty good grasp of the language actually.

说话人 5 02:07:13 
And I think that was quite successful. So I started like pharmacy books, updating those vocabularies, cool camera lists, and my Dawn and another Dawn. And now today, right, I'm at like 1,000 books that are published for more efficient languages. And one of it as well is create a website in our languages where I actually feature those books.

说话人 5 02:07:37 
Now this is what the website looks like for Cantonese books. You have the books on top. For each of the books, you have the paper decoration links, Amazon, for example, the ebook as well, so the popular ebook stores.

说话人 5 02:07:50 
And at the bottom I have SEO content. So I took part of the vocabulary list and publish them for free website. So Google could index it and people actually find it and use it and hoping that some of them will actually about the books. I did the model with English, but I did it for 30 languages because I realized if you had like English to German vocabularies, kind of also have to remember English vocabularies. So I only have to translate. They'd like to do the book and the content of the website. And I had the books for German, for Spanish, for the region, and some more languages. This is what it looks like for Dutch. You have the same layout just for the touchbooks. You will notice that the, actually the stores are different for that because the data has used different online stores. And at the bottom of the SEO content, again, in touch, but looks exactly the same as for English.

说话人 5 02:08:52 
So then get you into the data now. So for Google and for pain, this is what it looks like for 2025. I started the website 12 years ago. It's a quite an old website already. So yeah, 550 thousand clicks and the audience and 82 million impressions, not too bad thing is a very similar roughly one sense of what Google brings in.

说话人 5 02:09:19 
And when you go to countries, I actually use find something interesting. You would expect India, US, dangerous speaking countries above, but that's not the case. So Japan is about, they have roughly 120 million people. They're Germany with 80 million people, Italy with 60 million from us with 60 million attend already, Netherlands with only 80 million. Those basically one shenzh, Spain again, and then finance, we, Norway, all around under 10 million people. So a is probably similar, but it doesn't tell me. So they don't let it punch them all together. In rest of the world.

说话人 5 02:10:00 
And here's my period, small and rich countries, which are the ones I just mentioned, Norway, Herman, Netherlands, Sweden. I said you can ignore and I think it's too small, but I really like I sense included in what should I be common there is more exercise. So Shenzhen or smaller pretty much as the own language. Have a very high GPM capital, which is very interesting for you if you run assess good education and a good command of English. So back to the data. So why do they show up so high in terms of clicks? But I think just kind of speaking countries that are much lower than that is. So three potential reasons. Target keywords, broad keywords that make higher, or the brain simply don't matter in those countries looking at the data. So for Cantonese numbers, for example, for the page itself doesn't really help much in my case because in users actually show up, showing up a lot more than the smaller rich countries. So that's probably not the case for Bing. It's pretty much the same. You know, it gives me three languages. The other ones does, they don't have enough data. And if you go into the keyword itself, enter these numbers, again, the problem we follow that for too few links, after few searches, they didn't provide data. And Google Search Console for privacy reasons. So that's the case here.

说话人 5 02:11:38 
So target keywords, it's probably not the main reason why those, why does countries do so well. But if you look at here, then that's for the page, the key main keywords. So for English, you have long material keywords. For French, the first time we see that Cantonese actually shows up. So that's a very broad keyword. And then for that, we get into this very similar French. And if you look at this, so for English, the whole target keywords for French, you start to notice that like one broad keyboard pops in. But for touch, you have actually three role keywords in the top 5. So for keyword Cantonese, we have the same problem again with the clicks. It doesn't really help that Google is not providing that. But what you can see with the impressions is that actually, like for such a broad keywords, you would expect English to be at least 10 times, but it's pretty much the same as the smaller countries. So proportionally, it's showing up a lot more. If you look at the whole website as a whole and for the US market, you notice that on the top you have brain searches and then loaded on the target keywords. In Germany, pretty much the same branded searches and target keywords. But yeah, then for the Netherlands, you start, it starts with interesting where you actually have very broad keywords of ads. Means basically Arabic words, Korean words, Russian words, same for Swedish. What we can see from there is that for bigger countries like us, so bigger languages like German, the website ranks for target keywords, but it does not rank for broad category keywords. Whereas for smaller rich countries with less native speakers, it ranks for both. So for French numbers, but also for the categories. So for French vocabulary, French words, stuff like that.

说话人 5 02:13:59 
Finally, for branded searches, you can see Netherlands is fairly quite up at Denmark Suite as well. It's still like United States and social media. So the brand research is quite important for the English speaking market. But proportionally, Netherlands, for example, completely outbrings. The United States has three and 30 million people. Netherlands has 20. So you would expect like 50 times for clicks. Some of the data from Google and for Bing target keywords in country, we set the that the small and rich countries that like higher for broad keywords, it's a lot easier to ranked in. And for branded searches also showing up a lot more countries.

说话人 5 02:14:50 
I don't know, that's fairly far down the line for you guys because if you haven't started with Spanish or German, then those were probably coming out later. So those are the small rich countries. If you look at cheap people capital, nobody has 1990 thousand China is comparison to 20,000 roughly. So that's a lot higher.

说话人 5 02:15:15 
The next countries I would concentrate on are actually the former East European. So you need to be behind the Iron Curtain and like catching up now with the economy of Western Europe and potentially capital is still above China. And that was definitely worth considering if you're thinking about localizing another languages going down the road.

说话人 5 02:15:42 
Super sense. What does that mean? Process. So I took a couple of old ones, SMB, Oracle, Adobe, SMB of ones, and I look like what languages they actually localize for. If you look at the top, Adobe, SAP, Oracle, they pretty much have localizations for all markets. So the first part is the common languages like English, German, or East Japanese, Chinese, Korean. The second record is the smaller rich countries I mentioned. So maybe maybe Scandinavian countries. And the third one I have kind of ones, so the former Eastern European countries. So you see that the established ones, they realize localizing is worth it and they have a time. So they've been around for 10,20 years or longer, so they have the time to localize for the newer ones, shopify, Canva, Zoom, Atlas, and they mainly concentrate on establish kind languages, except for Canva, which is very interesting. Canva is a, B to C, Saas mainly. They also do product companies, but they try to get into these companies on the B to C level and then work the way out in there. So for them, they realize the SEO is actually a really good marketing channel in that way. And localization for them seems to work really well. So if you actually want to, they can take a case study, then you can look at them. And the newer ones, yeah, they simply concentrate on the establish ones and maybe Netherlands a little bit. So this is usually how sales goes. So you concentrate on the English market first, the hardest one for easier, then you go to German, French, Portuguese, Spanish, correct? Hardest ones. Then maybe you go to Chinese, Japanese, Korean, still quite hard. And then you form a population size because big population leads big target market, right? So you go with Italy, Poland, Turkey, Russia, bigger countries. And only then you start to maybe look at Netherlands or Sweden, countries like that. What I would propose, especially if assess, is that you did still do English first, of course. First and then probably German, French, Spanish, because if you don't have them, it looks like weird for western eye. But then after you do Germans, French and Spanish, depending on what size you have, especially depending on how expensive it is, you might actually want to start looking at the Dutch, Swedish, Norwegian, Norday's market because of their high GDP per capita, they have a lot more, they have pay power, so they actually have the money to afford assess applying the strategies or how do you go about this?

说话人 5 02:18:32 
There are four ways that work really well for me for finite languages. The first one is to use existing customers. This has advantages in many ways because first of all, if you already have existing customers in those countries, you know that the hurdles for them actually to use this ads. So Korea and as well as that, you can actually interview them. So you can use them to localize your sales for that country. You ask them, for example, what tools do you wanna see? What integrations do you wanna see? What payments, where payment members do we want to have? And based on that, you can in implement those and actually use those features in your copy for those language. So you don't simply want to translate the English content for, for example, Dutch, but you, for example, create custom copy for the Dutch market. You can do the English Royal Dutch directly and then which is in English and you translate into touch. This way looks a lot more c look more localized rather than like having a 1 to 1 translation of English content. For Google and website analytics is the same. You can see like which countries actually do already access your website and maybe use your sense. And then based on that, you can make decisions like which markets you end up next. If you already or if you're starting out, like actually localizing, so which most of you are maybe. So you have the English version already and then you go to German, Spanish, and French. You might want to think about using two or three landing pictures on your website to actually translate them in as many languages as you can. And then you end up maybe with 20 language versions for those script images. And then you get data. So then you let them sit there. You concentrate on German, Spanish and French, and after two or three months, go back to those, find the translations, find the languages, and look which one of those actually get traffic. And based on that, you can then make decisions like which markets do you have.

说话人 5 02:20:41 
Finally, free tools. Free Tools for SAS is like this possessed perfect application for free tools. If you look at hrefs, for example, they have and also free tools on the website because anybody likes free stuff. If you go about it, I would suggest that you actually take the free tool and put it on their own domain. So don't put it on your sales website, but put it on your own, on their automated, their advantages, they would be that you can then translate that application, the standalone application into many languages. Okay? And you get data like which markets actually need your services. So how to translate Google very quickly, because the other speaker already touched it.

说话人 5 02:21:32 
Can I help you with those languages? The answer is yes and no. So if you see on the left side, you can, you have the popular languages still CIS are confident they feel about translating into those languages. They're all pretty much are confident with any of the popular languages. But the further down right you go, the less confident they can. And what that means is that you probably, if you don't want to use it, like for the overall language designer, but you really don't want to use AI. So 1 to 1 translator website and just put it up for those smaller languages.

说话人 5 02:22:12 
So how would you go about it? I actually had really good success finding translators on fiber. Outlook is fine as well. Problem there is that some are really good and some are really bad. So that that's this is how I did the translation for the vocabulary list. I basically had able to 3 translators, then let them translate it. I go in an Excel file, transit, use 2 translate and 3 translate to see if they actually use that 2 one company 1 to 1. And based on that, slowly I could like find out which translator is actually good and which translators are bad. So you would think AI would help with that. But actually I think that makes that a lot hard enough. So if you want to look for translators before the backgrounds to simply use Google Translate and the technology was a lot worse. Now with AI, they can use AI and the quality gets to 90% there, but it's still not on the human level. But for you, if you don't speak the language, it's really hard to find out which ones are actually now using I and which ones are at least doing the work of editing AI and which ones maybe you can do it completed by themselves. Here, if you want to do this, my approach, I would suggest that you look for at least three translators, maybe 5,6, the more the backup, and then have to translate like maybe one page of your website. Then use AI actually to evaluate some translations. If you find typos, if you find grammar mistakes, take that as a positive, don't take as a negative. Because if you find out us, you actually know that's probably not star by I, it's probably done by human. And then based on that, you can go on and find the best translator and help them translate the content UCI to proofreading UCI to give you suggestions how to improve it. Me and then probably keyword translation, you can actually use AI for that. So the way I always went about that is you have the English keyword, ask AI to give you three or five translations of that, versions of that, and put that into Google. And then you will very quickly see which one is actually the best. So let's get sales data. That's a little more interesting, I guess. Yeah, the numbers of books I sell and the profit, if you look at the numbers of books, the smaller rich countries actually accounts for roughly 40% of sales. If you go to profit is roughly one third of the whole profit.

说话人 5 02:24:51 
But interesting here is if you remember the Google data, Norway, nowhere, Finland, Swedish, Sweden and Netherlands were pretty much the same. If you look at this, Sweden and Netherlands actually outperform by hello. So why is that? Definitely not the host? Because I localize for the country in particular, I have a, they're in the Netherlands and connected to all the sales channels. So all the online bookstores, offline bookstores. So if you want to buy my books, you can buy them anywhere in Netherlands. Furthermore, like touch people are very pragmatic, very down to earth. So they like her books because they're very dry books, very non non blah blah blah books. So they only contain the vocabulary if I go to the comments for Amazon England and UK or in the US, for example, everybody complains, there's no explanation how to do music. How do you actually like learning random blah, blah, blah. For that, I don't have this. So it's perfect.

说话人 5 02:25:58 
Problematic market feed. Something I didn't think about before, Jonas. I've just tried it and then I found out like that Netherland actually works really well for them. In Sweden, it's a similar story. So Sweden has similar sales numbers to Norway and Denmark until five years ago when Amazon decided to store for swing, a dedicated store for swing. Amazon of SE and then immediately like 5 x myself there. Similar story for Poland recently at Amazon and Belgium. So as you can see, traffic is one thing, but if you don't have the sales channel, if you don't make it convenient for people to actually buy. And this still more variables. What does it mean to says it means that you should, as I mentioned before, localize not just with translation, but we should localize with payment gateways. So, and also, for example, we have, I don't know, 10 or 20 payment gateways that I use per people. So depending on if you're B to C or to be, you would optimize for be different in making place.

说话人 5 02:27:09 
Same is true for accounting tools. Those countries usually have their own accounting tools. They don't use quick books or even zero. They have their own, their own basic accounting tools. Other tools like fulfillment is done with different tools. So if you can actually integrate those tools into assess, it will be a lot easier for other big other companies in that region to get on board. And we'll show them that you not only like translate the copy on the website, but you really actually localize down to features.

说话人 5 02:27:46 
Another strategy might be actually for Chinese companies to ride the wave of the big companies going abroad. So for example, JD dot com, just open the fulfillment center UK if you are SS doing working with e commerce companies, you could write that way and integrate into JD dot coms fulfillment center, something probably most UK based as don't have at the moment. Anything is the same thing. You could even maybe reach out to them and ask if you can collaborate with them, have them add you in the advertisement and things like that. All simply partner with local, like what's just their local players to get their tools integrated into assess.

说话人 5 02:28:33 
Some more takeaways from actually looking at kind of languages that might be available for you. The first one is PDF, actually, what would be beautiful research. So if you have anything that could be put into EDF, if PDF is a wine statement, a long blog post, or anything like this. Edit like once one, just want one copy into PDF and add it to your active articles.

说话人 5 02:28:58 
As you can see, people, I'm not sure actually if that's actually user's searching for it or if it's the eye searching for it. But I didn't dedicated searches for PDF. So whoever it is, people or AI, you wanna do it.

说话人 5 02:29:14 
Anyway, if you do it, then 4 things you should actually take care of. First of all, directly link to your website in the PDF. Because if you don't link to it and people don't know actually how to get back to yourself. Next thing is put the behavior, I think so as the website is so the same domain, don't put it on the CDN because otherwise the CDN UL will actually show up in Google. So you want to actually be beat on the same URL and be the name of the PM should be the target keyword of the blog post. And okay, that's pretty much.

说话人 5 02:29:56 
So free features is the next one. I already talked about that. So free features are great for SEO. They're also great to collect email addresses. But I would suggest if you do free features, make them as frictionless as well as frequent as possible. So don't require registration, don't require them to sign up for emails first. Because if you do this, then a lot of people would immediately go away, go back to Google and search another tool. And that would tell Google that your tool is not the one to actually list behind. So make it completely free. You can always like a user to sign UPS or banners or whatever later that. And last one is nicingan. So if your stress is like doing, I don't know, 30,40 things, it might be worth taking one or two of those things and getting them out into their own application. Already talked about that. You can do this for actually figuring out which market is it worth to go to. So get one feature, put it in a standard on application, and then translate it into other languages. But even for if you don't do it for like the main market, it's actually worth doing because like, for example, for me, I created a flash sale website after payroll languages because I saw that a lot of people actually look for flash ads. However. Searches for flash cards are roughly one tenth of that for vocabulary, so it's a lot easier to actually rank high. And by that, I actually managed to rank in for English really high for United States, United Kingdom, which I wasn't able to do for human languages. As a result, increase the sales from our focus for those markets.

说话人 5 02:31:45 
And now English books, I should celebrity, well, summary for rich and small language strategy that target countries with small population sizes and their own language, I actually worth targeting especially for high ticket items. So your assets are usually not the cheapest. So for you, those high GDP per capital and countries are actually a really good target market for SEO. It's a lot easier to rank compared to English or German or other popular languages. And you have potentially very little competition in those markets. So the big players, even the ones with a lot of funding, a lot of money behind that, they simply ignored my company until they are like 20 or 30 years old, those companies. And then they go into those markets. So for bingo languages, small language countries deliver 30% of 40% of the sales are roughly 1/3 of their profit.

说话人 5 02:32:43 
So I hope maybe in the future for your Saas applications, those smaller rich countries can do something similar. If you're interested, you can add me on Wechat. I also run, as I said, website service. So my main job is actually creating websites. It's not SEO. So if you have any other website targeting any of those countries, let me know. And yeah, looking forward to being in touch. Are there any question? English， Chinese or。

说话人 1 02:33:11 
大家先掌声好，然后再提问。嗯，有问题的吗？有问题举手啊。线上有吗？看看有没有线上想问的。没有问题，那我们就下一个。哦呦，是看到了，不好意思，又是David。

说话人 8 02:33:39 
Hi, I'm David. So just a quick question. Doing your, you know, looking through all these EOS keywords, what other, you know, industry or type of business do you see that this principle applies?

说话人 5 02:33:51 
Besides language, I guess it would apply to any industry to be honest, because if you look at Canva, for example, so they are very successful. They actually focus on SEO for English market as well. But if they think that those smaller markets are actually good translating into, I would assume that for them it's worth as well. The reason for doing it, I think, especially for SAS, is if like you are listening to me in English, like your English is pretty good, I guess, but listening to me in English is a lot more tiring than. So it's been Chinese, right? So it's the same for those people in those countries, right? They use it really good. But if you work with the software every day, you would appreciate like if it was in their own language. So if you can provide like a high figure, like a higher price Saas application in their own language, they would always prefer it compared to like other English on English only SS applications. Thanks.

说话人 1 02:34:51 
Oh, okay. Next, all of a sudden have to. Sure, lady friend.

说话人 10 02:35:02 
Great. Thank you. So would you mind sharing a little bit more about what's the difference between Chinese people to build a website and like western people are non Chinese people website. So is there any like the same process or development process different between?

说话人 5 02:35:19 
Yes, the definite differences. So the Chinese market is very different for SEO as well as, you know, it's same for websites. So anyways, the website is mainly there as a go to point for all the people here in China. Nobody really cares about websites in the west. If you're confident it doesn't have a website, then it looks very weird. That's the first thing that actually creating the website. The domains are very different. If you look at Chinese domains and western domains, they're completely different. We have very long domains in China, they're very short. And also, yeah, China, how do you say this? Western companies, they spend a lot more time, extra money on their website. So for them, it's a lot more important that they look good. They don't only take it as SEO, they actually take it as reputation for the company in China. It's more, it is, it's the opening at the moment, I think, I guess it will go get around to get more, especially the ones targeting western, well, western audiences, but boosted at the beginning of the transition.

说话人 5 02:36:35 
Great welcome. This was really interesting. I hadn't thought about localization, but kind of like the last two questions you gave us a B to C case study. I grant a B to B little agency not even assess like most put here what differences maybe you've noticed or could you just kind of riff on what I need to apply here in terms of the strategy you share for a small b to B service? It's, I think it's very similar actually, B to C and B to B in terms of SEO. I don't think there's actually that much difference because in the end, like normal people actually search for effect between the company and you search for a tool you actually use the same way as if you're at home. Like I search it for a book or something like that. So I think the difference is in terms of how you do it a lot that big. Of course, you want B to B one be more formal, maybe a little bit and more business focus on outcomes, whereas the B to C might be more emotional. Alright, actually, yeah, I'm not a marketing expert. So I think for that, but I think the difference I actually not that big. And yeah, does it answer roughly integrations?

说话人 1 02:37:56 
Okay, but later tonight, okay, we still have a few minutes if you want take more question or I actually have two questions, but this is supposed to put on the Q&A. But yes, as well, like 5 days, like I can ask, have you ever thought about setting your side project, meaning this website?

说话人 5 02:38:17 
Like, yeah, not really, actually, because I'm open to partnering with people. But yeah, I mean, like, if the right price comes along, additional sale, like everybody.

说话人 1 02:38:31 
But yeah, but once you worry, like what if something happen and then you don't make any money, any. More from this website. The reason why we're asking this is, I don't know if in this room we probably have some affiliate SEOs who got hit part in the Google average update in 2023 and March 2024. The reason why I'm like, oh, today we're making like grand. Tomorrow, next month cost will be 1 thousand and then next month right almost to zero. So in Chinese word we use that we. So the reason why I ask it is about the risk hint.

说话人 5 02:39:05 
I will hit this one. My driver used to be a little higher, but the good thing for me is like the books actually listed on all those marketplaces. So a lot of the sales actually come through the marketplaces directions, not through the website. So the website is basically there to, yeah, take it up and maybe get some sales. But I think most of my sales actually come directly to the marketplaces.

说话人 1 02:39:24 
Got you. Final question, do you know how much this project is worth? Are you comfortable sharing the number with us? Not sharing anything, but.

说话人 1 02:39:34 
OK，那我们大家掌声再次给。

说话人 11 02:39:43 
你的商业模式就是这个 ad 设计和这个订阅会员。

说话人 1 02:39:46 
有想学德语的，有学那个什么那个的。

说话人 11 02:39:49 
你的业务的话基本上是。

说话人 1 02:39:52 
大家觉得。

说话人 12 02:39:54 
算Saas，现在主要算 Saas 和订阅付费。

说话人 1 02:39:58 
要不等下 QA 的时候，大家问他们中文比拼一下，好吧？你。可以最后一个。你搞这个游戏上，你也。然后我们就休息了。那就是下一波人，赚一波人的钱。

说话人 12 02:40:11 
他天花板很低。

说话人 1 02:40:14 
收益很低。

说话人 12 02:40:15 
收益很低。接下来的嘉宾百万级、百万美金以上的。

说话人 1 02:40:21 
在过去这段时间，他尝试了一个 Microsoft 出海项目，最终虽然这个项目没有持续走下去，但在整个过程中他踩过很多坑，也验证了一些想法。今天他不讲成功经验，而是把这一次失败拆解开来，和大家复盘。他在出海过程中做对了什么？又在哪些地方交了学费？下面这个我就读不下去，我感觉都是 AI 写我的，我自己读也挺那个。好吧？那我们就直接开始，好吧？对，我就用这个翻页笔在讲，之前我想问一下大家有知道 Micro Saas 这个概念的有举手可以想举一下吗？真的这么少吗？有做 Saas 项目的吗？自己的做 b to c 的。好吧？当然有一个，我记得好像是，你是做 b to c 的是吧？做 b ToC Saas 的举手可以吗？就 c 端市场，而且他给的海外好像也不多。 b to b 的 b to b 有一些有。那我不知道大家是来自其他什么什么行业。

说话人 1 02:41:24 
OK，anyway，我这个待会讲这个Microsoft，好，然后有一天我收到了这么一封邮件，这前面是英语的，因为我们考虑到我们有国外的研究家在这里，所以我发了中英混合，好，大家理解一下。然后 2024 年如果没记错的话，6月初我收到这么一封邮件，当时我还没有看到，好吧？然后我就调查一下发信的这个人是谁，这是 AI overview 给我的答案basically，他讲的是这是一个美国那个 Meta Facebook 雇佣的一个律所。然后这个律所里面有人帮他工作，然后给我发了一封信，当时没看到这封信内容很长，但他的诉求很简单两个，第一个要移除的一个产品，这个产品是我们的产品。好，它是一款插件。

说话人 1 02:42:08 
第二个而且用了一个词叫 refrain from providing extension，就是以后你也不要尽量不要去做这类似的插件，意思有点像警告你，我如果再做你这个发在你同样的平台名下，有可能还是会再起诉你或者什么的好，然后他把里面的原因写的很多，有截图，大概原因第一个说他不会用我们的插件，好，第二个我们用了 Facebook 这个名字，在我们插件的名字里面好有侵权嫌疑。

说话人 1 02:42:34 
第三个，我们在调用 Facebook 的API，有爬 Facebook 用户的数据所违反了 Facebook 的 TMS TOS，好吧？发生什么事呢？这也是跟大家讲讲一个故事，然后总结一下我的这个经验教训，好吧？所以这个是我的，我们叫做 Micro SaaS project，因为这个很小，它不是一个像大家，像你们这个有四五百人的公司，做了很多条产品线，这种 Saas 也不像那个span，在那个我就不说名字了，就代表性的出版公司里面做那么大，好吧？然后几十号人弄一个project，没那么大，所以把它叫做 Micro SaaS，然后时间只持续了 8 个月，好吧。

说话人 1 02:43:12 
这是时间线，我们最开始的时候立项是 2023 年年底，然后最后决定退出是 2024 年的 7 月份，中间这是一系列过程，我就简单的聊一下这几个 milestone 第一个就是喜提的，第一个喜提就第一个用户评论一星的，好，这个是一星的，当时我看到以后要换麦，是吧？看见。

说话人 1 02:43:42 
对，第一个用户，我们这个插件上线以后，第一个用户来就说我们这插件。单独过了啊。而且说这个什么folder，什么什么之类，大概意思是他不知道怎么用好，然后我也去回复了，然后从这里开始我们就启动了这个漫长的迭代之路，然后后面又更多的异星评论来了，所以心里就慌了，我觉得这个怎么办？所以我就跟我的那个创始人，就这个项目的创始人，我就说这个能不能加快一点或者是重蹈再来？因为他实现这个技术的是用了一个什么框架，好吧？后面我们就说能不能用另外一个什么框架？用 react 点 js 或什么的。最后发现好像不太可行，所以我们就一直死磨在那个第一个那个框架里面。

说话人 1 02:44:22 
好，这是当时我们很快就拿到了 500 个下载，但是卸载率也挺高，可以算一下，基本上两个人下载一个，一个人就把它给删了。因为什么不会用？或者觉得没价值，好吧？所以大概率我们猜是不会用的，这是第二个这个 timeline 时间线就是比较positive，有用户会用的好，因为我们把产品迭代得好一点，稳定性也加强。

说话人 1 02:44:50 
我有一个五星评论，看到没有？所以我叫 v two，这个人我估计我一只要谈到这个插件，我就会记得这个人，这个 Lucy Adams 就说 v two is outstanding，给我们五星，好吧？然后我们还上线了一个兜内功能，目的是为了测试这个用户有没有付费意愿。

说话人 1 02:45:09 
如果别人真的觉得你这个有价值，我们又不想花太长的时间去开发一个付费功能出来，这个成本太高了，我们就测试一下，哎，看一下别人愿不愿意捐赠，好吧？然后我们也弄了一个这个，就在那个插件里面，当别人下载完把 Facebook 这个待会我会讲这个运营模式，好吧？下载完以后，别人 happy 了以后，我说，哎，能不能给我买一杯咖啡啊？然后那个 link 就会，这是一个link，是 point us 就链到我们的网站，然后那里有一个叫 buy me a coffee，是一个网站，有做Developer，应该知道，是吧？然后后来就收到律师团队给我们的那个邮件，然后从收到邮件到最后决定这个大概中间经历了一个半月，好，我们也没有放弃，但最后不得不关闭。

说话人 1 02:45:52 
好，所以下面会讲一些细节的。为什么我认为这是一个失败的项目，把失败又打上引号？有一个quote，因为这个我记得梁镇美在这里吗？是吧？我记得看过你一个采访，你说做了 100 个项目之后成功了一个，对不对？所以真的失败是习以为常的，所以我听完那个我觉得深有感触，就这一个项目起来 99 个就值得，但是问题是你能不能熬到？如果真的大概率事件发生，你 99 个都失败了，最后一个起来，结果你 98 个时候， 99 时候放弃掉，对不对？所以我觉得这个失败其实也不算是失败，我觉得也算是一个成功，因为他给了我很多这种启发，特别是教训。好失败是因为没有赚到钱。

说话人 1 02:46:36 
我们把时间搭进去了，好吧？我们都是技术出股，我管营销他，我另外一个搭档管技术，管这个开发，好吧？是营销方面和市场方面技术，然后也有一点小成就，就是获得了 1, 000 个用户，虽然最后留下来就是有人安装，最后又卸载，最后留下来只有应该是九百零几个，但我们最高峰应该达到了下载安装应该达到了1.5。k。

说话人 1 02:46:59 
好吧，在短短几个月内，大部分人都不知道 Micro soft，我还是挺吃惊的啊。这对 Micro soft 是什么呢？它不像是一个大软件就小微型的好，所以它也不一定是一个要本地下载安装，可以是一个插件，可以是一个APP。也可以是一个在线的工具，都把它统称为 Microsoft 的范畴，它目标，我们的目标是做一个微型的解决方案，快速迭代出来，解决一个非常利基，甚至是利基到三层以上的这种小型缝隙市场。

说话人 1 02:47:29 
好，然后把产品做紧也可以赚钱，而且这种欧美有这种开发大神或者是营销转去做开发的，也有成功的案例。好，我们这个 project 叫什么呢？名字？我就那个就叫 Facebook photo downloader，好吧？大家有用 Facebook 的举手可以吗？我大概调研一下有用Facebook。有更新飞书的举手，就你不仅用，而且经常发图跟朋友这种举手哇，这么少啊。还用飞书干嘛？投广告吗？那欧美有一个优特点，就是社交 Facebook 几十年， Facebook 上线应该是零几年还是06。

说话人 1 02:48:08 
そういった世代につなげるものだと、生産者の。

说话人 1 02:48:12 
我比如说我们当时遇到这个状况是因为我写了一篇博客，有一个人问我说我家人去世了，但他的所有社交媒体上的东西能不能把它下载下来成为回忆？好吧，他又不想注销账号，他也不知道他那个去世家人的飞书的密码，这是我们决定来，当时决定我去做这个插件的动机之一，好吧？所以这里面的场景就很多了，大家的账号密码不见了，回忆不见了，朋友 tag 你的那个照片不是你自己上传的，你也不见了，好吧？所以大家要的是回忆，大家要的是这个photos，所以我们就下载一款，我们就做了一款插件，是可以运行在Chrome，就是 Chrome 浏览器。那个 extension 好，然后用户点一下那个就可以点一下，我们发现选择那个 select folder to continue。

说话人 1 02:48:57 
我真的不知道有那么多欧美用户，真的，我不知道太笨了，还是我们这个只是觉得他点这里的时候他就不知道怎么选，其实我们要选就选一个本地的这个存储那个地址，来下载我们的保存他那个图片。好吧？很多人都卡在这一步，就点了我们插件，以为我们插件会很神奇，就一键把它飞书回来，你还没选地址呢，对不对？好吧？所以我们的用户有很多那个，这也是我们 UIUX 没有做好的原因，因为用户不知道怎么用的时候，他肯定给你打差评，他没有耐心去看教程，虽然后来也有了，这是我们学到的，这是我学到一个第一个很大的这种教训吧。然后点完以后，然后你就下载，你可以做其他事情。那个图片有几千张、几百张，我们都会慢慢慢给你下，我们当时一分钟可以下，好像是最快的时候应该可以下 60 张，当然这个比不上我们竞争对手，但我们是免费的。

说话人 1 02:49:46 
好，然后你可以下，你的后面应该还有可以下你的图片，你可以下你朋友图片，可以下载任何一个，那个基本上只要那个能爬取商业模式是什么呢？这就是一款 Chrome extension，可以下载用户的 Facebook 图片，而且可以保存起来，是怎么获取的目标用户，当然很自然的，我们就说 Facebook 所有用户都是我们的目标用户。所以当时一看到这个数字， 30 亿，全球基本上 40% 人都在用Facebook，都是我们的目标受众。哇，这个市场好大，所以就犯了第一个，我们觉得太贪心了。好吧，你不能定位所有人，特别 c 端，然后还有什么动机？什么失去，什么账号密码，什么被 Facebook ban 掉？或者是家人去世了，朋友去世了，他想把回忆拿回来，很多使用场景。

说话人 1 02:50:34 
好，我们解决什么问题呢？就是我们为什么做创建市场？有什么产？有什么解决方案是不能满足的？就是看一下 Facebook 自己有 backup 的这个功能的，但是 Facebook 是不想让你离开的，所以你可以想象一下一个公司怎么会把一个备份方案做到极致上的回忆。

说话人 1 02:50:53 
一下子就拿到自己，然后以后他就不能从你身上榨取任何价值，所以用户对他来说就是价值，你就是他产品，所以他那个 backup 功能很烂，你用过就知道，你要等今天 request 点下载，你要等到明天或后天才能收到，你打开以后一个包，结果那个包里面都是一些 profile photo，还有几张你最近的那个像素也很差，好吧？所以他没有动力把它做好，所以但用户需要这个的，然后当然就没法去下载，我要只下载那个文件夹、那个album、那个相册，或者下载朋友太多的图片，或者下载我的什么的，他没法指定好吧。

说话人 1 02:51:26 
当然还有第三个就是用户是很忙，大家不想一张。去下载，你想过去十几年重度用户可能就有几万张图片，对不对？最主要原因就第三方没有好的完美的解决方案，就大公司觉得这一块做起来真的很简单，但是鸡肋市场为什么？因为那个 Facebook API 经常变，他一变他那个软件就失效了。然后他又没不想去投入一个小团队，专门去开发这个插件出来，万一别人不付费或者付个小钱，所以这一块的话是我们看到过我觉得最香的一块。就大公司知道这块市场存在，也有技术能力，但没有动力去开发，好吧？他也没有动力把它做好，所以已有的方案就是飞书这个下载功能只是它产品的一部分。

说话人 1 02:52:09 
我的那篇文章在我后面会介绍，就是我当时推荐了几个软件。就可以实现这个功能，但那个软件有 100 个功能，只是一个功能负责的，所以用户会下载这个到你的 PC 或Mac，然后再去下载那个，下载它飞书图片可以做成，但是要这过程那个 user experience 不好，然后我依然拿到佣金了，为什么有人愿意付费？好吧，所以这是我觉得这个钱一定有人愿意相付，已经基本上算是验证过了，然后第五个那个图片的这个质量问题，好，时间关系我就不多说，所以我觉得一定会有一个更好的方式，这是我们的这个设想解决方案，我们也出来了，我们把它叫做 game 前景了。好吧？当时确实是很兴奋，觉得很inspire，觉得这个肯定可以整。然后我们提供什么核心功能？ one CLICK back to school，一键下载好吧。还有你可以随便选你这个想要下载的，而且我们会无损，没有那个图片下载都是按照你这个 1: 1 的，没有说把它压缩下来，这我们提供的价值获客也想清楚了。

说话人 1 02:53:11 
最简单的就是修改一下我们已有的两篇文章，我有两篇文章，当时只要去搜 down Facebook 都有 down Facebook APPS 之类的一连串的这种，那个基本上我都排在前三。

说话人 1 02:53:22 
好吧？我有两个站，我是做内容站长出身的，然后长期的话等我们这个插件出来，我也可以当成这个品牌，我去获客，我去投广告，好吧？这是设想，刚才提到这是我一个网站，当时那个流量还可以，一天就关于这一这个话题的，就别人搜的，就我就几十，有时候高的时候 100 个流量进来，好吧？ 100 个用户在搜这个，他们都有这种诉求。然后我排名还可以，当然这个是 GSC 的，我排在一，基本上排在前三组，关键词还有附带一些长尾也会看的啊。然后我看那个 top query，这个站长应该都知道 GSC 的怎么变现，我们有三个变现方式，freemium，就是有一个直接性对手叫个easysite，好，那个是一个公，是一个美国人，一个程序员，所以这是为什么？我觉得我们可以打败他，因为他就一个人，一个程序员，我说我们也有一个程序员，还加一个我有市场，我说我们肯定可以搞定他，是吧？但我们低估了他这个人也挺厉害的。就他提供 200 张免费下载， 200 张以后你要付费，他把那个 Saas 模型都跑通了。 200 张多少张？你最高可以付 100 多刀， 100 多刀，对于这个下载 Facebook 这个已经算是怎么说？应该是一个什么一个比较高的价格了，对吧？就看你这个张数，就是 300 张付 100 乘以个零点多少，大概收你个 20 块钱美金，好吧？所以我就说我们就 300 张甚至 500 张，我肯定要内卷把他卷死，这是我们的设想。

说话人 1 02:54:50 
好，然后第二个 value add，就是我们可以做这个增值服务，我觉得那个程序员肯定没有想到这个好，增值服务就是用户要的真正的是，要的是那个记忆，那我们把记忆不仅仅把它下载保存那里，我还可以打印成相册。给他留作纪念，是吧？我们可以跟那个相册公司合作，把这个例子，把这个导给他，然后把这个产品寄给用户，这是我们设想好第三个最重要就是为什么叫Saas？就是 software subscription，我们可以卖服务，就是你依然在用Facebook，可能还有人想要备份，还可能还有那个自动备份需求，所以我们就说等你有朋友带给你，你只要用我们这个软件或插件，以后我们可能涉及到所有图片自动备份到你云盘， cloud storage，你想那个的都可以，这是我们的设想三大市场潜力，这是我刚说 30 亿用户，别人都说只要拿到 1% 就行，我说我不要1%，我 0.01% 都可以。好，你可以想象有多少是我们的目标，那个受众，好，我们只想赚一点小钱，我们就两个人，不求大富大贵，这是我们的设想，但是这个一切都成为泡影了。后面会讲为什么会追求这个三个原因。我刚才说了，我有资源可以推广它。

说话人 1 02:55:59 
第二个我有一个直接竞争对手，我可以去分析逆向去 reverse engineer，好吧？基本上这个没问题。第三个我还有个搭档 1 + 1，我一定比那个人更强、更厉害，对不对？而且这个竞争对手叫Linkedin，你现在还在，如果有人还想做这个产品，你可以跟我交流，我可以给你很多idea。好，他现在还在那个 hootsuite Facebook。他有多少个用户？七万个注意，不是。7万，7万的还在用的就有人下载，然后可能付费了，最后把它给删掉，这个都不算这个数字，好吧？7万，好吧，我们的计划是什么？这是我跟我搭档那个合伙人写的一个计划，三个阶段，第一年做什么？一到两年做什么？然后长期的话再做什么运营思路都有了，团队分工了，什么基本上都解决了，但是你知道有时候就人算不如天算，叫做我们真的没有想过会收到 Facebook 的律师函，为什么？因为我们竞争对手还看到吗？就他赤裸裸的把飞书商标都放在这里， TM 都放这里，而且他可以存在这么多年，好吧？我觉得如果遇到了我，就算遇到我也不怕，因为这次我依次为这么干了。

说话人 1 02:57:09 
大家就说他没，就是飞书为什么盯我？他应该盯头部，就是枪打出头鸟，我们肯定又不是出头的，所以我们就不担心这个，等到他来的时候，我发现比我们想象要复杂一些，更去 handle 一些，好吧？然后一个半月具体细节我们就不，我们中间也没有放弃，来来回回跟飞书说我们把这个去掉，我们承认错了，我们又没有这个要侵犯用户隐私的权益，什么什么之类的。

说话人 1 02:57:36 
来来回回，来来回回，一个半月后发现还是不行，那我就最后给我搭档发了一封信，我说不行就下线算了，因为我们还有其他的Saas、 Micro Saas 也在做，这只是我们的项目之一，然后他也同意了，所以这个项目就完了。好吧？后来当然没有时间去反思，我们依然在做这个 Saas 项目， Micro SaaS，但后来我就有时候会去晚上时候，这个时候我就想，诶，为什么 Facebook 会可以找到？我们？怎么这么厉害？而且我们只是一个刚刚上线才几个月的，这个产品用户量也不多，是吧？所以我就复盘了一下，因为这个对我们这个计划。

说话人 1 02:58:10 
来。太快，它是一个，怎么说？是一个致命性的一个一个事件，第一个就是我们的核心产品功能确实不够不稳定，就是不管是Windows、 Mac 或者是那个其他那个操作系统上用，其实多多少少都会出一些问题，好吧？这个跟我们这个技术的这个框架有关系，而且我没有重蹈再来的这个实力竞争对手，那个 Eclipse 确实比我们做好，那个程序员就是很牛。好吧？我也问过我那个业界那个程序员，那种大牛有的还搞不定。

说话人 1 02:58:40 
这个虽然听起来简单，第二个就是用户体验。不加，就你看很多一星评论人用的飞书，前面有可能有人还不小心给飞书去投诉，或者飞书就他的肯定会有人去跑，他一定会有个公关团队，特别这种大公司，你要用他的域名做什么或者是什么 TT 叉，然后用户体验不佳，他就觉得你的产品的一个体验，好吧？还有一个原因就是上线太早，这也是给我们不管以后这个 AI 时代，以后来就不要急，你把产品做好，就不一定要做的很烂，就觉得他们是测试市场，最后不行，其实这个理论已经失败了，我告诉你已经过时了，因为这个不好的用户体验， AI 就是放大镜，很多人都说你不好，哪怕你这个产品在最近三个月发生了翻天覆地的变化。

说话人 1 02:59:26 
但是一年前、两年前 ready 说你不好， Google cross DOMAIN 用户说你不好，这些东西都成了 AI 的这个引用源，他说这个产品依然不好，所以他会 discourage 用户去下载你，他会引来一些不必要麻烦，所以你上线产品，你的这个产品，这个甚至 V1 的不行，可能要 v 几什么那个迭代几个版本，真正觉得稳定性了再发出去，好吧？这是给我们不管是 ToB 还是 ToC 这傻子创业者的一个一个lesson。

说话人 1 02:59:53 
我觉得好做对事情，因为只有 2 分钟了不到了，我就快速过一下，我觉得既有做对的，也有做错的，我们就看一看就好了，没做好的这是我的一些可以跟大家分享，就第一个不要太贪心，第二个你的 position 一定要。很精确，我们定位了全世界所有飞书用户。好吧？觉得都是的，但是不是所有下载飞书的人都愿意付钱，都愿意有那么耐心，所以一定要细分下去，找到你的那个理想客户画像，也就是我们前几位嘉宾说的 ICP 这个过程有了 AI 的加值会很快，但是很多公司很多，哪怕是刚才丹尼尔江森也提到，就是很多这个硅谷那边这个很牛的人，那个公司 santa 什么模式的他也不清楚，这个过程非常重要，就找准你的ICP，你找准你的 ICP 下面所有的营销 SEO 真的就会很明晰，好吧？好，其他我就不啰嗦了。

说话人 1 03:00:47 
好吧，这是经验教训， take away，我看能不能快速耽搁两分钟。我讲一下，因为很重要，第一个要做 Saas 的就是小团队和个人，你不要去碰那个 b to c 了，真的 b to c 获得用户很容易，但是他也一个坏用户，就是这个现在这个时代就是说一个什么东西不好和你这边去处理这个评论他是不对等的，你说一个东西不好一分钟搞定，但是处理这个有可能花一个小时还不止，好吧？这就是为什么很多大公司就没有那个 PR 资源去处理这些公关，你不处理，但不处理它会影响以后那个客户对你这个信任。

说话人 1 03:01:21 
好，所以尽量不要碰 Microsoft 这个 b to c 方向，你就要专注做 b ToB 好吧。我在清迈，可以多啰嗦一下，我清迈一个 SU 好友跟我一样内容站被打了一塌糊涂，在 24 年，当时他比我就先退出的好，中间又去做UZ，他做 UZ 应该做了一年多，结果最高才一个月只赚 2, 000 美金，但是 AI 从 20 我们就半年不见，我觉得清迈就不怎么那个，他也不说半年时间他转行来做这个 AI SaaS，他跟我们很多那个我不知道戈飞在不在这里，戈飞的朋友圈有没有人在这里的？有 follow 戈飞的吗？就跟你们一样，他也是从零开始去学做那个。

说话人 1 03:02:02 
做那个AI、SaaS，或者是那个什么有一个新的东西出来做一个网站，就那个，是吧？他半年时间，他现在告诉我，我不能说他名字，他我们去蒸桑拿，我们就喜欢汗蒸，蒸桑拿，他就说建蒂说我有个新项目，就几个月时间起来，我现在挣到一个月 15, 000 美金，你想到 Facebook 当时做那个 Youtube 的时候，一年多才挣了 2, 000 美金， 1.5 k 就十倍的差距。

说话人 1 03:02:28 
好吧，他为什么能在这个Saas？而且他就一个人，而且满世界跑，没有团队。他的团队是什么？他说他就用那个 cloud code。这是他的团队出来，然后这个人有，我们没有什么，他有 SU 经验，他懂德国市场，他有德国人，他全世界旅游，他知道就这个市场的那个需求能够把搬到另外一个市场去，他很厉害这方面，因为这是我们 SU 人那个人发现痛点，然后他有了这个，我们把他技术给使用之后，他给我一个案例，因为我们两个比较信任，我看了以后我觉得这个要割碎的朋友圈来说，他花一天时间，我说你们可能花两个小时就搞定了，真的就这么简单，但他能赚到钱。这个你看过他网站你就知道，就是真的很简单，就是这里面就发掘那个机会。

说话人 1 03:03:11 
好了，他只做 b ToB SaaS，他不碰 b ToC，你看他定位的这种 Saas 用户是非常标准的。我打个不恰当，比方就是比如说深圳的殡仪馆，好吧？他们在去服务客户的时候出现什么问题？我们能不能给他提供这种 Saas 服务？你可以朝着这种方向去，谁会想到这个利基市场，对不对？好吧？然后第二个，这是问我们刚才问那个 Wolf 店的时候，我说你没有离开策略，我们当时没有任何离开策略，就是当你去创业时候的创业项目。一定你给自己留后路，不要留后路没问题，但你一定要有一个离开策略，就是当你这个项目因为什么东西做不下去，团队上火或者飞书这事件来以后，你处理不了，但你之前做的东西依然是有价值的，你会对谁有价值？你这个东西可以卖给谁，可以转手给谁，这就我们所说的离开策略，你可以卖掉，你可以转让掉，但这个东西都有价值。

说话人 1 03:04:02 
最后我一分钱没捞到，为什么第二个我们没有想好离开策略就是满腔热血，我们就开干，好吧？就去做，但这个插件被移除了以后没有一点这个value，而且我没有赚到钱。我们也，我们连网站还没起来，好吧？我们刚还在内测这，于是。没有一个成的，一个品牌，第三个没有 all in 的决心，真的很难成功。你不要小看一个Microsoft，就它对你的这个要求还是挺高的，好吧？就然后第四个这个一样可以解决，就不说了啊。然后我个人觉得 Saas 启动成本越来越高，虽然有了 AI 加持，技术方面门槛越来越低，但对你上创始人、创始团队的这个商业的这种嗅觉，还有对客户的理解，对竞争对手的分析和你自己对这个整个市场的感知，其实是一个很大的，有很大的挑战。所以这也是为什么我经常要出来办这些活动。

说话人 1 03:04:51 
我说大家不要一直闷闭门走，也不叫闭门造车，就闷在头里，你要去看世界，为什么我喜欢去？去研究那些地区首脑都在讨论什么，因为这些地区首脑就是世界的这种风向标，哪个国家、哪个地区发生了什么，有什么机会，这是第一批最先知道的。好，所以你去清迈是世界那个地区首脑应该是排行前十的，前五的都有可能，你在那里，他一个全球人都在那里做什么东西，有什么那个一下子就知道了，这是为什么？我们要鼓励大家出去，要出去看你出不了国你就出去，出来社交总有能拿到一些信息，好吧？这些信息为以后你的项目一定会奠定这种有利的这种基础，哪怕跟你现在做项目无关，这些技能在哪个点一定会可以出现出来，好吧？然后不要孤独创业，一定要找一个互补的搭档，这是我很久以前就明白的一个道理。

说话人 1 03:05:42 
最后一定要检查邮箱，我知道国内人不喜欢检查邮件，有时候等你的，我当时就错过了一个多星期，我要早一点的话，或许我的想法也不一样，所以等到没看到时候，而且飞书我还发来 follow up 那个律师团队，所以发来 follow up 的时候我们就慌了，就有点好吧，所以一慌后面这个节奏就不对了。你节奏不对，有可能你就往着更坏的方向去想，所以有时候创业过程就是一个节奏的过程，就像你打球一样，为什么有时候不是我喜欢打篮球，说比谁投的准，比谁把节奏控制得更好。

说话人 1 03:06:12 
好，OK，最后我叫虽败犹荣，所以我们现在这个插件页面是这样子的，你点 URL 进去，所以 this item is not available，好像这个就不存在，其实我觉得是依然有回忆好吧？所以这为什么我要利用这次分享会把它分享出来，证明这个项目虽然已经失败了，但它至少我们有奋斗过。好，就这么多，谢谢大家。下一个应该是茶歇，应该是，对吧？好，那我们就茶歇 15 分钟。好，外面应该还有那个吃的吧。有的。好， 15 分钟以后我们回来，好吧？

说话人 11 03:06:57 
你，你们最开始是批量搞站，现在是批量搞这个插件，搞这个应用。

说话人 12 03:07:03 
一直都是做网站。

说话人 11 03:07:04 
做网站。那然后一个代码，然后放一些工具在上面。

说话人 12 03:07:12 
都是，也是，写代码称不上都是群友。

说话人 13 03:07:17 
对，都是群友，我是那个听，然后你们听说过。

说话人 12 03:07:22 
好几个，好多群，之前我不知道我们线上聊过没。我们线上聊过吗？

说话人 13 03:07:27 
线上有啊。

说话人 12 03:07:29 
哦，那应该我有，我看到过，我有印象，那个印象。

说话人 13 03:07:35 
你不是问我那个，我还以为你问我呢。

说话人 12 03:07:38 
我以为。

说话人 13 03:07:39 
我之前去年那个分享会的时候见过你，但是太火了，知道谁都找你。算不上，就是没机会。算不上。

说话人 12 03:07:51 
没机会找你，现在都没怎么出门了。

说话人 13 03:07:54 
现在确实没是没时间出门，忙着赚钱，是吧？

说话人 12 03:07:58 
哪现在都是天天烧TOKEN。还真烧TOKEN，花不完就焦虑。

说话人 13 03:08:07 
慢慢做嘛。 200 刀就能花完你现在 TOKEN 用的。怎么样？就是每天大概多少消耗？

说话人 12 03:08:13 
一天几个亿吧？

说话人 13 03:08:14 
一天几个亿，好家伙。我们是 500 兆的会员。他两个号吧。 2 个号， 3 个号， 3 个号，两个号都不够他造了。

说话人 12 03:08:31 
多花点钱，多花点TOKEN，反正又不招人， 25 个Codex，你看 25 个，对吧？

说话人 1 03:08:43 
你就跟监工一样，你就天天看着他们干活。

说话人 12 03:08:48 
25 个 Codex 我看一下。

说话人 13 03:08:50 
有一个公式。

说话人 1 03:08:51 
就是你们肯定不用那个养虾，你就直接用 Codex 干全部的活。

说话人 12 03:08:54 
一样的虾，手机上养的手机上其实没有电脑效率高。

说话人 13 03:09:00 
那你虾，你是用 CC 吗？

说话人 12 03:09:03 
都用 Codex cloud，都用，今天4亿，昨天8亿，前天 15 亿，那明天5亿，哈哈哈。

说话人 13 03:09:11 
这一般都分布用来，应该不都是用来写代码吧。

说话人 12 03:09:16 
都是写代码。都写。我现在不是都写，是什么？都用写代码，当然写代码多一点。

说话人 13 03:09:23 
你同时开发好几个项目。对啊。那我现在最近发现 Codex 用了两下不好用了，我感觉它。

说话人 12 03:09:34 
比较慢，稍微慢一点也还好，我觉得都挺好用的。

说话人 13 03:09:40 
主要是我让他去交外链什么的都有点费劲，他就不执行，你知道不听话？他说违反什么政策什么的，有没有碰到过？

说话人 12 03:09:52 
没有，你违反什么政策啊？你用的 Codex 吗？我现在用的Codex，我之前用 CC 额度一会就没了。我 200 多的 CC 我是用半天就用完了。

说话人 13 03:10:02 
我是用不动，我是没这个钱了，但是我就是用自己的那个 PPT 账号，然后给他授权登录到。 Codecs GB GPT Codecs 5.5 点几？5.2。

说话人 12 03:10:15 
我现在用5.4， 5.4 好用， 5.4 好用，你看我们现在。

说话人 14 03:10:21 
5.4 GPT 吗？

说话人 12 03:10:22 
Codec 我用了 25 个号， 20 个号都用完了，昨天一天 4 个亿， 8 个亿。

说话人 14 03:10:28 
你们现在用 Claude code 用的多还是用 cursor 用的多？

说话人 12 03:10:32 
一起用 Claude 的话， Claude 一下额度就用完了。我用两个号，我两个 200 刀的。

说话人 13 03:10:38 
你开发应该用。

说话人 12 03:10:40 
25 个 Codecs 就是一块用我，因为我浪费的 code TOKEN 比较多，我又不懂代码，然后我就要反复的就开一堆终端嘛。这个核对一下。

说话人 14 03:10:53 
你是让他们核对，是吧？你用其他的扣点工具是为了让他核。

说话人 12 03:10:57 
这都是终端，这是Codex，这是clawcode，两个小时就用完，然后我现在就是做了一套就是换号的切号的 25 个Codex，这个用完了就切下一个自动切嘛。自动切，然后这个闲鱼上 6 块钱一个，我买 200 个。

说话人 14 03:11:19 
也就两块， 1, 200 块钱。

说话人 12 03:11:21 
1, 200 块钱，那我能用很久很久了。 200 刀都不够用。

说话人 13 03:11:26 
500 刀都不够用。诶，他这个应该他现在。

说话人 14 03:11:28 
在闲鱼上买的号。

说话人 13 03:11:30 
CC 不，他不是那个很严格吗？

说话人 12 03:11:34 
用 Google play 花 3 多花 30 美金，他就没有，他不会封号了。礼品卡，是吧？正常 200 美金，我是 270 美金的，他应该对，正常 250 美金还是 270 美金，还有 20 美二三十美金的税， 50 美金的 Google play 的费用。

说话人 13 03:11:49 
诶，等一下，他都用礼品卡了，那你为什么可以？

说话人 12 03:11:53 
我没有用礼品卡，我直接国内信用卡付款的。礼品卡你还得买什么呢？我就信用卡付嘛。

说话人 13 03:11:59 
很贵，那。

说话人 12 03:12:01 
对，太贵了。那你 Codex 质量很好， Codex 用 5.4 Codex 的好处， Codex 团队版钱有很多， 6 块钱一个很便宜。 Codex 我买 20 美金跟 200 美金的，用的烧的很快，太快了。

说话人 13 03:12:19 
充 20 块钱，然后 90 刀一天的额度。

说话人 12 03:12:24 
哈喽哈喽。哦，大佬你好。大佬你好。我刚看到你上个月的分享，我刚才还在看呢，好厉害呀。

说话人 13 03:12:34 
谢谢。

说话人 12 03:12:35 
哪 2 月份冠军还是 1 月份冠军？

说话人 4 03:12:37 
没有，不是冠军。

说话人 12 03:12:38 
拒绝，是吧？

说话人 13 03:12:41 
全都大佬，哈哈，彪哥更是大佬。

说话人 12 03:12:45 
我现在 verso 的费用好高。

说话人 14 03:12:47 
吴达标。

说话人 4 03:12:49 
今天晚上的晚宴。

说话人 13 03:12:52 
面基来了，我我我我刚刚问飞哥了，飞哥他说他家里有事。

说话人 12 03:12:57 
他应该没有来。

说话人 13 03:12:58 
不然要面基一下的。我们现在做了多少个站了？几十个吧。我记得。现在他不是。

说话人 4 03:13:16 
做不来了，我是一个人做，但是就感觉又要运营，然后又要做。

说话人 13 03:13:23 
这是哪位？

说话人 12 03:13:25 
刚才群里分享的非常奈，高飞群前两天不分享一个万刀的吗？

说话人 13 03:13:32 
记起来了，名望。

说话人 14 03:13:42 
招员工会好一点。

说话人 1 03:13:43 
招员工。

说话人 14 03:13:43 
比较简单，你要找一个合伙人的话，他不是要找个人结婚。

说话人 14 03:13:57 
最好是招一个，或者你招的话最好是找猎头招。就是你要招一个，比如说举个例子，合伙人是 90 分，你找一个 75 ~ 80 分的给他工资开高一点，他可能也相当于半个合伙人的角色，但是他就是那个啥，就是给他工资嘛。

说话人 4 03:14:28 
就是看他工作经验有多少年。

说话人 14 03:14:35 
那你要是这个领域要有七八十分，怎么也得个三五年以上吧？你高级嘛？你就想当于你去招一个高级嘛？

说话人 4 03:14:42 
架构师那种吗？

说话人 14 03:14:43 
用架构师。

说话人 13 03:14:44 
用你座位吗？

说话人 1 03:14:46 
不是。接 API 就行。

说话人 14 03:14:49 
这很简单，你就找一个高级的就好了，你就相当于公司有高级产品经理，是吧？有高级前端、高级后端、高级这个营销人员，是吧？相当于你是找一个高级嘛？

说话人 14 03:15:08 
不需要让他用AI，或者可能写得比我还慢。

说话人 4 03:15:13 
万一写得比我还慢怎么办？

说话人 14 03:15:15 
写代码，写的写代码。那肯定不会嘛，他无非就是补充你不会的那个趴嘛。

说话人 13 03:15:23 
要求说清楚，感觉要求说清楚。

说话人 4 03:15:27 
你看我像程序员。

说话人 13 03:15:31 
程序员得是我们这种头发稀疏的。

说话人 14 03:15:36 
说法叫古法编程。

说话人 13 03:15:39 
古法编程。

说话人 4 03:15:40 
古法编程。

说话人 13 03:15:42 
然后。古法手作。

说话人 14 03:15:43 
不用。不用 AI 的这个程序员古法编程。

说话人 13 03:15:48 
又到 Web coding，然后现在叫什么engineering？古法手艺。那有用那个虾胶外链吗？

说话人 4 03:16:01 
不行，我不能要。

说话人 13 03:16:03 
你有用龙虾胶外链吗？

说话人 4 03:16:04 
抢不上。

说话人 12 03:16:05 
有，我今天群里发过，发一些简单的，但是系统的发还是得哥分那一种，你得大量的发，龙虾发的有点慢。

说话人 13 03:16:13 
然后你在用的时候就完全没有说他，他会涉及到说他觉得这是违反什么规则，违反什么规范，他没有说他就直接执行了，就你怎么说他就直接执行。没有。

说话人 12 03:16:28 
你的说法不太对，你是什么搞什么批量的，那个 chatgpt 的道德感是高一点， Claude code 的道德感到 Claude code 会干。是这样， chatgpt 它总是会触发。就是什么说这不能干，那不能干，是这样。是这样，对吧？是这样。

说话人 13 03:16:47 
我还以为就我的有问题，

说话人 12 03:16:48 
是这样，不过我是两个混合用的，我智谱也买了。我智谱也在一块用智谱的，太慢了。智谱的，唉，效果太差了，还是跟现在是 5.4 还是跟 Codex 5.4 和 Opus 4.6？就这两个。

说话人 13 03:17:04 
Codex 4 点和 5.4 它不就是 GPT 吗？它应该也会触发一些。

说话人 12 03:17:09 
是这样，是有一点，但是。它的代码，它的紧密性比 OBS 好， OBS 还是会遗漏一些东西。

说话人 13 03:17:18 
因为我在想，就是我现在一调教不动codecs，所以我打算换mininix。

说话人 4 03:17:24 
也有。

说话人 13 03:17:32 
刚刚才知道居然 Reddit 也能买帖子。

说话人 12 03:17:37 
都可以买它，是它那个策略可以买。

说话人 13 03:17:42 
还是可以搞一手，因为目前来讲是我只涉及到特色博客这些，还是。

说话人 13 03:18:00 
梁总，你现在买博客吗？做个博客。

说话人 12 03:18:03 
还没有到那一步，我现在在弄品牌站，咱从头开始弄。

说话人 4 03:18:09 
我有，在，我用。他全新开发了一个，就全程他开发，然后帮我上线一个APP。

说话人 13 03:18:17 
让虾开发吗？

说话人 4 03:18:18 
对，全程了，好像是我就没有动过，就是他弄上去， push 上去之后我在手机上看看。

说话人 14 03:18:25 
哦，上架了吗？

说话人 4 03:18:26 
在审，应该过两天因为被打回来了一次，再让他。

说话人 13 03:18:32 
我不太敢，我怕他改什么东西不告诉我，哈哈哈。

说话人 1 03:18:37 
我养的那个虾就是太多的，网上找一些文章给他。然后他看这个又看那个，看这个学到后面有点混乱，不知道他架构到底调成什么样？

说话人 13 03:18:48 
而且上下文长度有限。

说话人 1 03:18:51 
还是没有一个很好的文章，就是能让他一步一步，或者说他只学一样，而且他学了很多东西，走火入魔了。

说话人 14 03:18:58 
把 HR 和 Cplus 的文章都看了。

说话人 12 03:19:01 
我等于把它全部抓下来，给它做知识库。对。是的，全部抓下来。

说话人 4 03:19:08 
然后还有让他提交外链，然后发到网站。

说话人 14 03:19:12 
让他写文章，写很长的文章， informative 的文章。

说话人 1 03:19:17 
写了。

说话人 12 03:19:18 
各做各的。就一个su，一个文件夹，上面一堆的skills，我让他写了，就是 pulify 三木哥的那一套嘛。直接给 Youtube 的API，然后抓所有视频转文稿，然后再去写，就这一一个 Su 文件夹，就相当于一堆马克龙文档交给他，让他按照这个干就可以了，然后再配上。

说话人 12 03:19:39 
有疑问吗？反正我之前没有，我对比了一下，有一点点流量比没有强，而且我不用干啥，然后这个可以复用，我把这个文件夹放到另一个网站的文件夹。 Banana 自动配图。就是你得迭代它的scale，优化它，让觉得哪不行，让它改一下，很快这个东西就是消耗TOKEN。

说话人 4 03:20:03 
我也做，我是做了个插件，对，我没有直接让它。

说话人 12 03:20:07 
就相当于所有的知识库都是一个文件夹，所有的文件夹都可以分发，然后他就让他开多个窗口让他干就好，所以就每天消耗，整体消耗每天 10 亿到百亿的TOKEN。

说话人 1 03:20:21 
多干这个活，一个文档被读 n 次。

说话人 13 03:20:26 
对。怪不得那 feelgood 你用的是哪个模型？ CC 啊。你用CC，然后。虾也是用CC。

说话人 4 03:20:36 
Max 那个套餐。

说话人 13 03:20:38 
哦， Max 套餐够用吗？ 200 刀那个？

说话人 4 03:20:40 
还行，不行？再买一个吧。

说话人 1 03:20:43 
可以。不是买，不是问题，关键是封号。

说话人 4 03:20:46 
我还好，我的没有被封。

说话人 13 03:20:48 
没有被封，是吧？

说话人 4 03:20:49 
因为我还用美国信用卡。

说话人 13 03:20:52 
哦，那主要是他就不让授权，就是说最近他不是说不让授权用到别的那个，比如说 agent 这种项目之类的，他都开始风控了，好像。

说话人 1 03:21:07 
我都还好。最后合影。抽奖是抽。是 Lark 高级票，是吧？两张那个深圳 SU 大会的那个。

说话人 11 03:21:18 
现在这个高级票能帮你们选品吗？

说话人 1 03:21:20 
如果没记错，你看我们官网应该是 480 元一个，可以参加连续四天的活动。

说话人 12 03:21:25 
其实它跟外贸。

说话人 1 03:21:26 
好，我们要等到。

说话人 12 03:21:30 
杂活、小活肯定都可以。

说话人 1 03:21:31 
好，最后一个演讲嘉宾，我们要再次请回这个Danielliang，OK，前面他给我们讨论了流的增长的系统、流量、社区以及 ready 以及市场选择。但对于 Saas 公司来说，创始人本身的号召力。也是不可忽视的变量，这一点如果大家有去考察过北美、欧洲的一些 Saas 公司或初创公司，会发现他们有个特点，创始人都很active，都很local。

说话人 1 03:22:04 
好吧，好像我们这边好像不一样，跟我们的文化和环境有关系。当时丹尼尔问我说要不要做这个演讲的时候，我说台下应该没有人对这个会感兴趣，但是我依然想让你讲出为什么，我想让大家见一见外面那个 Saas 公司的创始人一般是怎么来做这个 IP 的，怎么利用自己的 personal authority 来提升他的这个跟他 list 这个项目的关联，好吧？这其实也是marketing，也是 growth 的一部分。所以我们本着想让大家开开眼界的这种角度，也让他再继续讲一个好。而且还有一点，就在 AI 内容泛滥的当下，真实的人和公信力会变得更有价值。好数字人还是不如线下这种人好。丹妮会在最后这场分享中拆解如何将创始人的权威感转化为可拓展的搜索这个需求引擎。大家再次欢迎丹妮江神为我们做压轴分享。 TT need cool.

说话人 5 03:23:08 
I hope you guys had a good lunch.

说话人 1 03:23:11 
I appreciate it's been a long day. You've been spending a lot of time listening to a lot of cat speakers. I've taken a lot of information. I also feel very dumb because I thought I knew this field and then people speak and I'm like, wow, I didn't know a lot. So one of the things that I talked about in the first session was basically how AI, it can be inauthentic. You know, I'm sure that, as I said, when you go on LinkedIn, you scroll through, click the top , you know, it doesn't really add value. You don't learn from it. You don't take something that's gonna increase your revenue.

说话人 1 03:23:44 
So what I wanna talk about today is founder led distribution. And essentially what this means is how do we utilize the people within the company to promote a genuine voice at scale. And essentially that's what I'm gonna be looking at today. And raise your hand if you already do any of this. I'm kind of curious.

说话人 1 03:24:08 
Founder led content production. Okay, interesting. It's only the speakers at the back where I got away. So I think who's heard of the phrase build it and they all come. Yeah, for those who haven't heard it, maybe it's not a phrase in China. It's essentially this concept of if you have a really good idea and you go to your friends, your family and you speak to them and you say, hey, I've got the best idea in the world. And then you explain the idea to them. They'll then say to you, yes, dear, that sounds fantastic. You should definitely do that. So what happens is you put your blood, your sweat, your tears into developing something, and you think it's fantastic that you think it's gonna change the world. There may be lots of money, you're gonna be able to buy an island in Hawaii. And then what happens is you create this, you launch it onto the market and nothing, you don't get customers, you don't get engagement. And you're confused because, you know, you spoke to a bunch of people about the idea and they all said it's fantastic. And when you released it, nothing happened. Why is that the case? And I think that's kind of the first item here, invisible. We're in such a competitive space at the moment that it's really difficult to differentiate ourselves, and I think it's also really difficult to niche down because we don't wanna exclude potential customers, right? So that's challenge No. 1. Challenge No. 2 is there is so much money in this space, right? There's so many investors. Everyone wants to be the person who invested in the unicorn in the open claw or if it was a private company. And as a result, it's really difficult to launch your company when your biggest competitor spending a million a month on acquisition to your target market. And I think this one is actually one of the biggest challenges.

说话人 1 03:26:19 
Generic positioning messaging. As a dyslexic person, I find, but I initially found very boring. I never really investigated it because I was more excited about platforms to market than I was the actual messaging positioning. And then I realized that having effective positioning and being able to align that with your customer is arguably the most important thing that you can do. So the challenge that people have is they launch their product, they use maybe AI to create content to promote it, and he still don't get that engagement. And one of the biggest challenges that, or one of the biggest opportunities that I think we're seeing is the opportunity for breaking through the noise by having genuine, authentic, real content. And one way you can do that is through finally distribution. And the reason why is it kind of goes back to my first session, when I was talking about how the customers at SAP didn't care about 0.4% increase in efficiency or 0.04% recent efficiency. What they cared about was the personal problems that the product was able to solve for that, the fundamental fraction. And that's because when we look at brands, when we look at companies, we often personify it. And it's really difficult to remember something if it's generic, if we don't care about it, if it doesn't relate to us in any capacity, which is why I think it's such a great opportunity now. When the world is just filled with AI slot, to be able to use an authentic voice. And it's also a lot easier than I think people think it is.

说话人 1 03:28:14 
So the first is, founders have real insight. Now every founder I know thinks that they know more than everyone else in their entire industry about the product. So you have to be a bit careful about that. But for founder gonna spend the effort to create a company, there's obviously something behind the scenes that they know potentially other people don't. People follow people. So I think if anyone here is promoting content on LinkedIn pages, do you actually tell me people's profiles do much better, right? And that's because people don't really care about a LinkedIn profile and people connect with human beings.

说话人 1 03:28:56 
And then also like developing a relationship with your customers takes time. You can't immediately assume that people are going to connect with you, understand you and appreciate you. It is something that you have to build consistently.

说话人 1 03:29:12 
So this is kind of the process that we follow internally and we scale starts, which is No. 1, kind of content. So identifying the types of content that actually work does take time. What I like to do is develop a spreadsheet, look at different value propositions based on the transcripts that we're having with customers and like performing research online. And it's important not to exclusively focus on necessarily one topic because you can often be pigeonholed within that space.

说话人 1 03:29:42 
Next is thinking about trust and authority. What I mean by that is, why should we care about you? What is unique about your experience that's gonna add value to our process? Why should we listen to you? And then inbound conversation. So that's thinking about who we're trying to talk to and what are the relationships that we have at the moment. Combine all of that, and you have a process that you can follow again and again on a renewable basis.

说话人 1 03:30:15 
So I love LinkedIn. I hate going on it. It's a lovely relationship because every time I gotta, I get really angry because there's just so much rubbish on the platform. Everyone's like, I discovered the secret. Just type in 10 x into the comments and get my free PDF. All my god.

说话人 1 03:30:33 
I, even if I see something that I find interesting, I still don't do it because I hate the idea of indulging people in, in their attempts. But what we've been able to see is a lot of effect, raised millions for various customers just from organic content on LinkedIn pages alike, LinkedIn profile pages. So function increase in engagement cross channel. So the biggest issue is often if you no one cares about your product in the way that you can, right? You're the biggest fan of your company, of your client, right? And it's really difficult to objectively look at things. So you have to make sure that when you're publishing something, will anyone care if it's not gonna add value, if it's not actually going to generate your response or engagement, you probably shouldn't do it. Don't cares about those random updates that don't add any value. Don't cares about the Christmas party that you have company. No cares about the insights that you learn from marrying your partners. What people do care about is what controversial takes. They do like that. But I also find that if you do too many controversial takes for the sake of being special, no one's gonna care about you. They do like genuine insights within the industry, and they do care about people that are developing the product.

说话人 1 03:31:59 
Bear in mind, with a few different things that, so the way that we look at it is, No. 1, we develop a specific post, we make sure that it's aligned with our BP. We make sure that we're thinking about our messaging and positioning. Ironically though, even though we spent all this effort, some of the best engagement has come from ideas where the founders left the shower. Voice noted it in a WhatsApp to me. I've converted it into a post, published it and go back engagement. Once we've posted it, the next stage is getting them to engage with it. And once they engage, we can find a way to reply to them. We DM them and do all of this automatic. And the last thing we do is we learn. And what I mean by learn is we develop a playbook, and it's in motion. And it's essentially every time we publish content, we add into this, we take away the learning. So it's, this is the length of content that worked. This is the type of topic that engage people most effectively. This is the time of day, etc, etc. And the goal isn't to be perfect. The goal is to develop learnings over a series of time, so that we can increase the accuracy and alignment of the content that we publish with our customers. So based on the book that we, the, based on the playbook that we've developed, these were the four learnings that I extracted actually for this specific presentation. So, and this is all from a revenue perspective, like in terms of content that's actually generated revenue.

说话人 1 03:33:36 
No. 1. If you're not enthusiastic about the post and you don't think that it's actually going to, if it's one of those generic AI spot posts, it's not gonna work. It's better to spend a week thinking about something good and actually working on something that's decent and provide some level of insight rather than making sure to hit Monday, Wednesday and Friday.

说话人 1 03:33:59 
Data works really well. Obviously, numbers work well, but having a narrative is arguably the most important aspect of all of that. People connect the narratives, they build that relationship really well. A lot of people miss out on the replying aspect of it. People leave comments and then those comments go unanswered. This is a great opportunity for you actually to engage with your audience and inconsistency. So what I was trying to say there is kind of like a lot of people are like, I need to publish 10 posts this month, but rather than doing that, make sure to post on a regular basis. If you can't do 52 posts, which obviously is, it's a lot, I would always say go for like one post a month and make sure that it's really valuable. I'm curious actually for a Chinese market, who uses Linjin here region? Okay, interesting. What's the Chinese LinkedIn? Yeah, lingy.

说话人 1 03:35:10 
Every time I download my media and it goes to that website and then I have to. Oh my God, interesting. I thought maybe like xiaohongxu or something was how people use the area. So capturing insights, some of the most valuable, there's a few different areas that you can do here. First is shower thoughts work really well, surprisingly. Next is you can extract insights from the conversations you're having with your customers at the moment using stuff like Reno, no AI or Slack even. I would also encourage you to get other members of the team to contribute insights. So for example, with one of the healthcare company that I'm working with, they find that actually the best content has come from customer service. So the customer service opposed on Slack saying, we spoke to this person, this is the challenge that they had. Maybe we should talk about it. So ideas can come from anyone. You can structure it. Be careful about how much structure you apply to it. Like, because if you apply too much, it comes across as rigid and it doesn't feel engaging. And then definitely the best important is distribution. I think people post and then expect other people to come. So what I'd recommend is that you develop some sort of like checklist. And then in that checklist, you have a series of different communities and people that you know will engage. So for one of my clients, every time we post, I say to them, hey, can you post this message or this comment at this time? And most of them are usually very happy to do so. I do find that you use the same people to engage in your post overtime, those people's impact reduce. So, and I think what happens is that LinkedIn is seeing, okay, this guy is probably encouraging these people to engage. But I think it is really important to make sure that you consistently post afterwards another idea, another each way that you can promote your article is by including it in the newsletter. And then that means we can duplicate content as well. So the areas where I'm finding this most useful, we get a lot of it in Geo, which I'm a bit surprised. But apparently some of these algorithms are actually reading LinkedIn. I think also because anyone can publish any content now, the real differentiator is the quality of quality as it has always been. But I think a lot, but people, it's difficult to internalize that. And then obviously community and Reddit matter regardless of the industry, even if your enterprise, these platforms are not to be missed. Okay, so here's three of clients that we're working with.

说话人 1 03:38:21 
Correct.

本文档为站内渲染。原始文件本地路径：saas/source/seo/SEO-seo方法论-ppt-SEO-John-3e09e2.txt（仅本地保留，不入库不部署）