FoundConf Mike King.pdf

THE AI SEARCH
PLAYBOOK
IT’S NOT JUST SEO.

Michael King

Founder and CEO
iPullRank

Search Engine Land Just Gave Me This Award
for the Second Time

THE
PROLOGUE
SOME THOUGHTS BEFORE WE GET STARTED
4

I RECENTLY THREW
MY FIRST
CONFERENCE
Andouille capicola doner
picanha, quis flank
exercitation pork belly
chicken. Jowl biltong
porchetta beef incididunt
sausage doner t-bone ut
mollit andouille. Aliqua
magna occaecat quis short
ribs beef ullamco. Salami
drumstick pork loin ea.

https://ipullrank.com

5

YOU KNOW WHO
ELSE PERFORMED?
ME.
This was the moment that I
decided I don’t want to have
to win over crowds anymore.
I want people to already know
the music. So, I decided I’m
going to make a hit record.

AI Search Playbook |

…SO I WENT AND
WORKED WITH
SCOTT STORCH
A couple weekends ago, I flew
to Miami to work on a song
with the guy behind hits
from Beyonce, 50 Cent, Justin
Timberlake, and many more.

SCOTT STORCH
HAD HUMBLE
BEGINNINGS
Scott grew up in Philly too
and was an early member of
The Roots.

AI Search Playbook |

SCOTT STORCH
SCORED A $100MM
FORTUNE
The guy used his piano skills
to go to the top of the A-list in
the aughts.
Get it: “scored?” 😏

HE ALSO PRETTY
FAMOUSLY LOST IT
ALL
He had a very
well-documented downfall.

AI Search Playbook |

It was one of the craziest days of my
life. I was around a bunch of people
who believed they could do anything
at anytime…and did.
— Me, CEO, iPullRank

AI Search Playbook |

Now is the moment to bring all those big
ideas to life. There is nothing stopping you
from doing the absolute coolest thing you
can think of and changing the world around
you.
— Me, CEO, iPullRank
AI Search Playbook |

THE FUTURE
OF
SEARCH
IN BRIEF
16

SINCE 2016, I’VE SAID THE FUTURE OF SEARCH

LOOKS LIKE THE MOVIE “HER”
At HubSpot’s Inbound Conference I explained
that we had all the tech required to make the
assistant from “Her” real.
Just last year OpenAI rolled out their version of
ChatGPT as a voice assistant. Now that they have
memory and other integrations, the technology
in the movie is effectively real.

PRAGMATICALLY, THAT PLAYS OUT LIKE THIS

Weʼve all had the experience of 50 open tabs and the shopping cart we forgot we filled.
The proactive search experience alerts you when you are passing a store with that
product.

THE FUTURE OF SEARCH IS AGENTIC
THE AGENT COLLECTS INFORMATION FOR THE USER PROACTIVELY
NOW

NEXT

Consumers discover, interpret, and consume
information.

The agent does more than browsing. It interprets
and prepares the information for the user. It
delivers a ready-made, cohesive answer.

Discover

Interpret

Consume

Discover

Interpret

Consume

PAGE

19

THE GAME
HAS CHANGED

Google is still the main stage for discovery, but
search behavior is fragmenting.

AI Search platforms including AI Overviews (and AI
Mode), ChatGPT, and Perplexity, are shifting the
way people search.

Visibility now depends on presence across all
search stages.

PAGE

20

SEARCH HAS GONE FROM DETERMINISTIC TO PROBABILISTIC

Deterministic Search

Probabilistic Search

What you put into the search engine came
out the same on the other side and the
results were consistently replicable.

Results are augmented through reasoning
chains and the output changes based on
an array of variables.

ANSWERS
COME FROM
EVERYWHERE

Conversational Search stitches passages together
from various sources based on invisible
subqueries.

These AI Search systems donʼt simply rank URLs.
They retrieve, blend, and synthesize ideas.

The more surfaces you appear on, the greater
your chance to be included. You win the raffle by
holding more tickets than the competition.

PAGE

22

PASSAGES GO THROUGH PROBABILISTIC “REASONING CHAINS”

We have no control over what makes it through the reasoning chains.

“PERSONAL CONTEXT” IS A FUNCTION OF VECTOR
EMBEDDINGS

Google is creating embeddings about all of a userʼs interactions to filter out results. Consequently,
embeddings are the technology that fuel search. The prediction thereof is what underpins Generative AI.

VISIBILITY IS
MULTIMODAL

Success now requires more than optimized text on
your website.

Discovery happens through text, video, product
data, podcasts, PDFs, and social.

Your brand must meet users in every format
and on every surface.

PAGE

25

RELEVANCE LIVES
IN THE OVERLAP

Classic SEO relies on crawlability, indexing, and
authority.
Conversational search depends on context,
knowledge quality, and synthesis.
Relevance Engineering unites both, giving your brand
the machine-level advantage to stay retrieved and
recommended.

The brands that engineer relevance will
define the winners of AI Search.

PAGE

26

THE FUTURE
OF SEO

Search is strategically different from what the
industry and software can currently support.

Winning means creating content that resonates and
building persona-driven, audience-first matrices
mapped to the new search journey.

Relevance Engineering is the framework that
makes sense of this complexity and keeps
your brand in the answer set.

PAGE

27

USER BEHAVIOR HAS EVOLVED AND SEARCH IS
BIGGER THAN GOOGLE

WE’LL HAVE LESS CONTROL OVER WHAT THE USER SEES

DUE TO HOW USER BIASES

Every interaction the user
has with the system
biases the output in more
meaningful ways than
we’re used to with search.
Brand efforts will have
much more of an impact.

IN SOME SPACES THERE ARE DIRECT
INTEGRATIONS THAT BIAS SEARCH

Zillow isn’t just a listing site anymore, it’s becoming the “ask-and-see” portal for home search in the AI era. For
consumers, this means simpler, more intuitive access to homes. They’ve created integrations with ChatGPT and
Google and that appears to be biasing results in Gemini and AIOs. Those with relationships with ChatGPT will
likely benefit. Google is likely to remain less bias here.

SEARCH IS BECOMING MORE OF A BRANDING CHANNEL

Even if these results donʼt yield a click, your brand is now a part of the userʼs consideration set from this non-branded search.

ADDITIONAL SEARCH MODALITIES ARE MATURING

Users have gone well beyond their desktop computers and are searching
in the real world. The technology has matured and will continue to evolve
to the point where the line is blurred between the digital and the physical.

THE FUTURE OF SEARCH REDUCES FRICTION BETWEEN THE DIGITAL &
PHYSICAL WORLDS W/ A PROACTIVE ASSISTANT THAT UNDERSTANDS YOU
Users never
wanted 10 blue links.
They wanted the Star Trek computer.
The Future of Search is a multimodal
user experience that is far less about the
interface, and users synthesizing a
wealth of information.
Itʼs far more about the evolution of
technology helping users do more of
what they want in the real world by
developing a long term understanding of
individuals and giving them exactly what
they want before they ask for it.
Brands need to be proactive to keep
meeting their users where they are at.

SEO
IS
DEPRECATED
IT’S NOT DEAD, BUT IT’S NOT WHAT I DO
34

AI Overviews are Changing User Behavior

AI Overviews are Causing Traffic Losses
Wikipedia, one of the
most Organic Search
reliant websites saw its
traffic grow with the
introduction of ChatGPT,
but has seen its traffic
decline with the advent of
AI Overviews.

STUDIES SHOW
AIOS REDUCE CTR
BY 34.5%
A recent study of 300,000
keywords by SEO platform Ahrefs
found that click-through rates
drop by 34.5% when Google
displays an AI Overview. This
confirms that the impact of
AI-generated answers isn’t
limited to a few websites—it’s a
systemic shift affecting the entire
search landscape.

I LOVE TO SAY I TOLD YOU SO

I PREDICTED THIS 2 YEARS AGO
https://searchengineland.
com/how-search-generat
ive-experience-works-an
d-why-retrieval-augmen
ted-generation-is-our-fu
ture-433393#

— Somebody in Mountain View, Google
AI Search Playbook |

TRAFFIC AND CONVERSIONS HAVE DIVERGED

AIOs Yield Less Traffic, but More Qualified

SEO IS HAVING A
BLOCKBUSTER
MOMENT
Back in 2000, Blockbuster
executives reportedly laughed
Netflix out of the boardroom
when offered the chance to buy
them for $50 million.
Fast-forward to today. Netflix is
worth over $200 billion, while
Blockbuster filed for
bankruptcy in 2010.

AI Search Playbook |

TRADITIONAL
SEARCH IS
MELTING DOWN
The foundational assumptions that
built entire marketing departments are
crumbling:
Rankings = Revenue — Irrelevant when
users never click to your site
Content = Quantity * Keywords —
Useless when AI systems synthesize
information from multiple sources into
a single response
More pages = More traffic —
Meaningless when AI answers
questions without sending users
anywhere

AI Search Playbook |

CLICKS ARE
DECLINING
Sure, Google still holds the
lion’s share of overall user
activity and search traffic as
of Q1 2025, but the landscape
beneath our feet is shifting
dramatically. Recent industry
analysis shows
year-over-year growth in
clickless queries while clicks
to organic results continue
declining.

LLM TRAFFIC IS
MORE VALUABLE
Semrush is projecting that
the value of the classic
Organic Search will
continually drop while AI
Search channels will gain
grown dramatically over the
next four years.

AI Search Playbook |

WE MISSED THE
MOMENT. IT’S GEO.
While our industry spent the
last year in-fighting about
what things should be called,
the market has named it.
It’s called “Generative Engine
Optimization.”
Let’s move the fuck on.

AI Search Playbook |

WHAT WE DO AT
IPR IS RELEVANCE
ENGINEERING
(R19G)
“SEO” and “GEO” never made sense
to me. We were never “optimizing”
search engines. We were feature
engineering inputs for search
engines.
Ours is a channel-agnostic approac
that is is the confluence of AI, IR, CS,
UX, and DPR.

SEOS NEED TO
EVOLVE
Modern search visibility requires
engineering relevant content that
can easily show up in every type
of search platform. This is where
GEO comes in. GEO builds on
similar value systems that
advanced SEOs, content
marketers, and digital PR teams
already understand, but the
execution requires entirely new
skill sets, tools, and strategic
thinking.
AI Search Playbook |

INTUITION WON’T
CUT IT IN THESE
ENVIRONMENTS
The days of SEO “best
practices” based on educated
guesses are coming to an end.
When you’re competing for
visibility in systems that
process billions of data
points, intuition isn’t enough.
You need statistical rigor.
AI Search Playbook |

RELEVANCE
ENGINEERING IS
GROWING BEYOND
CHECKLISTS
SEO is a “checklist culture”
where people mindlessly follow
outdated best practices, which
is quite wrong. Instead,
engineering for relevance
requires establishing a series of
technical standards rather than
relying on anecdotal evidence.

NEW ROLES ARE
EMERGING
The functionality of these
new channels has evolved so
new skills at the intersection
of AI and content strategy
(content engineering) and
deeper understanding are
required to navigate them.

YOU NEED TO
UPSKILL TO GENAI
The adoption numbers tell the story.
According to Stanford’s 2025 AI Index
Report, 78% of global organizations are
now using AI in production; up from
55% in 2023. This surge in adoption is
creating massive hiring demand for
new skill sets.
Generative AI skills in job postings
have exploded across every category,
with some skills seeing over 2,000%
growth year-over-year. Prompt
engineering roles increased 350%,
while retrieval augmented generation
positions jumped over 2,000%.

AI Search Playbook |

NEW SOFTWARE IS
EMERGING
If ChatGPT’s Atlas browser
catches on, it’s likely going to
change the user interaction
model of the web even more.

We are moving from a Google-shaped web
to an agent-shaped web.
— Me, CEO, iPullRank

AI Search Playbook

THE
STATE
OF
AI
SEARCH
TRENDS AND ALL THAT STUFF YOU CAN SHOW YOUR BOSS
53

USERS NEVER
WANTED 10 BLUE
LINKS
A user wants answers, but
classic search has high
access, cognitive, and time
costs that make the process
too inefficient.
Google’s has gradually
shifted to an answer surface
rather than a portal to the
web.

AI Search Playbook |

User preferences, and what users
want, is also changing. So, instead of
factual answers and 10 blue links,
they’re increasingly wanting
contextual answers and summaries.
— Markham Erickson, VP , Google
AI Search Playbook |

SEARCH JOURNEYS
ARE MESSIER THAN
EVER
While the search journey has
never been linear, the user
navigates the stages in a
stages in an even messier
back-and-forth now.

OVER 50% OF GOOGLE SEARCHES RESULT IN ZERO CLICKS
SOURCE: SPARKTORO / JUMPSHOT
As Google has emerged as an answer
surface, more than half of users are not
clicking through to results the open web.
This has continued to accelerate with the
advent of AI Overviews.
However, users that click through are more
likely to be ready to take action.

ZERO CLICK
SEARCHES TO NEWS
SITES ROSE TO
67% YOY
Based on data from
SimilarWeb new sites went
from 56% zero click to 67%
YoY. Publishers have been hit
the worst by these changes.

AI Search Playbook |

ONLY 1% OF USERS
CLICK LINKS IN
AIOS
According to a study from the
Pew Research Center user
behavior swings dramatically
when AIOs are present with
users clicking out to the open
web 47% less and ending their
search session 62.5% more.
Note: Google has refuted this
study.

USERS QUERIES
ARE EVOLVING
SLOWER IN GOOGLE
Although AI Mode and
ChatGPT are similar in
function and users give more
robust and freeform prompts
to ChatGPT, they are still
ramping up prompt length in
the Google environment.

HIGHER QUALITY
CLICKS FROM AI
SEARCH
Adobe’s recent study has
indicated that the gap in the
conversion rates between AI
and non-AI channels has
closed significantly in a very
short time frame.

…BUT A NEW
GERMAN STUDY
SAYS OTHERWISE
🚥 <0.2% of traffic vs Google’s
organic
📈 Conversion rates rising, but
AOV falling
🧭 Great bounce rates ≠ deep
engagement
Source:
https://papers.ssrn.com/sol3/p
apers.cfm?abstract_id=5585812

AI Search Playbook |

AGENTIC
COMMERCE
PROTOCOL WILL
CHANGE
E-COMMERCE
OpenAI’s and Stripe rolled out
a specification that
revolutionizes shopping feeds
and allows users to transact
directly in generative
environments.

52% OF ADULTS
USE LLMS
A study by Elon University
found that 52% of U.S. adults
now use LLMs, 34% of whom
use them at least once a day,
and that number is sure to
grow.

AI Search Playbook |

AIOs HAVE GROWN
RAPIDLY
Semrush’s analysis of over 10
million keywords found that
AI Overviews were triggered
for 6.49% of queries in
January 2025, climbing to
7.64% in February (an 18%
increase), and jumping to
13.14% by March – a 72%
increase from the previous
month.

THE GREAT
DECOUPLING
More impressions: Your
content may be seen more
often because it’s being
summarized in the
Overviews.
Fewer clicks: But users may
never visit your site because
they already got what they
needed from the summary.

THE BIGGER DEAL IS HOW IT SKEWS THE SOURCE OF TRUTH
GOOGLE JUST CHANGED THE DENOMINATOR
Like it or not, GSC is the universal source
of truth for what is happening in the
Google ecosystem. That source of truth
effectively just changed the denominator
in its calculations and now everyone
thinks they have more visibility than they
do.

GOOGLE IS HIRING
SO IT WILL GET
WORSE
Google has two job openings
for Anti-scraping analysts so
you can expect that rank
tracking is going to get a lot
trickier.

AI Search Playbook |

Protip: How about we stop targeting
keywords with 30 MSV when we track
them daily in SERP tracking tools.
— Me, CEO, iPullRank

AI Search Playbook |

REFERRAL TRAFFIC
IS LOW FOR THESE
CHANNELS
From May to July 2025,
SimilarWeb data indicate that
fewer than 20% of Google
search users click a link to an
external site, while fewer than
4% do so in AI Mode. This
sharp gap in outbound clicks
from AI Mode signals the need
to completely redefine KPIs for
measuring organic search
performance.

AI Search Playbook |

REFERRAL TRAFFIC FROM CHATGPT AND PERPLEXITY
HAS GROWN 4X SINCE 2024
Data from SE Ranking shows that AI traffic
from ChatGPT and Perplexity has grown to
0.13% globally – four times higher than in
2024.
It’s still a rounding error for most websites,
but the value of the traffic cannot be
ignored.

If you’re making decisions based on just
traffic, you’re missing the point of the
channel.
— Me, CEO, iPullRank

AI Search Playbook

GOOGLE IS STILL
DOMINANT
Google still drive 8X more
traffic than ChatGPT and
90% of users that use
ChatGPT use Google.
Google still drives more
referral traffic than any
source.

AI Search Playbook |

Many new platforms will emerge, but
Google will be the long term winner.
— Me, CEO, iPullRank

AI Search Playbook |

GOOGLE’S
PROPRIETARY DATA
ADVANTAGE
One of Google’s biggest
advantages is its access to
proprietary data at scale.
While many companies rely
on publicly available content
to train their models, Google
taps into its enormous,
constantly updating stream
of data.

FIRST-PARTY CHIP DEVELOPMENT (TPUS)
GOOGLE DOESN’T NEED NVIDIA
While many AI companies rely on
third-party chips like NVIDIA (which is
suffering from a chip shortage), Google has
built its own Tensor Processing Units
(TPUs).
TPUs are custom-designed chips built by
Google to accelerate the kind of math
operations that deep learning depends on.

BILLION-USER
PRODUCTS
Google owns multiple
products with over a billion
active users, giving it an
instant audience to test,
deploy, and refine AI features
at unmatched speed.

AI SLOP IS
GROWING FAST
A Stanford University study
analyzing over 300 million
documents (from corporate
press releases and job
postings to UN press
releases) showed a sharp
surge in LLM-generated
content following the launch
of ChatGPT in November
2022.

AI-GENERATED CONTENT IS 16.5% OF GOOGLE SEARCH RESULTS

According to Originality AI, AI-generated
content accounted for 19.10% of Google
search results as of January 2025, up from
just 7.43% in March 2024. Even as that
figure dipped slightly to 16.51% by June,
the signal remains clear.

AI USAGE DOES NOT
IMPACT RANKINGS
According to data from
Ahrefs, 86.5% of top-ranking
pages now include some form
of AI assistance. Only 13.5%
are fully human-written.
Google appears largely
indifferent to this shift. That
is, it neither significantly
rewards nor penalizes
AI-generated pages.

HOW SEARCH
INDEXES GET
SYSTEMATICALLY
POLLUTED
In April 2025, Ahrefs analyzed
900,000 newly published
English-language web pages
and found that 74.2% contained
AI-generated content. Only
25.8% were classified as purely
human-written. While 71.7%
were categorized as a mix of
‘pure human’ and ‘pure AI’.

AI Search Playbook |

REDDIT CITATIONS
HAVE FALLEN OFF A
CLIFF IN CHATGPT
Speaking UGC, according to
data from Profound, Reddit
has dropped from a peak of
12% of citations in all
responses to below 1%. In
recent weeks it has climbed
back up to near 3%.

AI Search Playbook |

90% OF ONLINE
CONTENT MAY BE
AI-GENERATED BY
2026
Digital Ouroboros is on the way with
experts estimating that as much as
90% of online content may be
AI-generated by 2026.
For creators, marketers, and businesses,
this raises urgent new questions. How
do you compete, rank, or even stay
visible in an ecosystem buried under an
avalanche of low-quality synthetic
content?

AI Search Playbook |

THE SEO HEIST
The 2023 SEO heist story is a
classic example. In this case, a
co-founder openly bragged
about using AI to “steal” 3.6
million visits from a
competitor by scraping their
sitemap and generating 1,800
articles at scale. The
problem? All those articles
were published with minimal
human oversight, prioritizing
pure volume over any
semblance of quality or value.

AI Search Playbook |

HALLUCINATION IS
NOT A SOLVED
PROBLEM
Vectara’s hallucination
leaderboard shows that
accuracy improvement
remains modest: GPT-5
achieves a 1.4% hallucination
rate.
Gemini’s 2.5 Pro has a 1.1%
hallucination rate.
AI Search Playbook |

THE MISINFORMATION PROBLEM
STUDY REVEALED AI SEARCH IS WRONG 60% OF THE TIME
A March 2025 study by the Columbia
Journalism Review tested eight major AI
search engines and found that, for a
controlled information retrieval task,
chatbots collectively provided inaccurate
or misleading answers more than 60% of
the time, nearly always without
acknowledging uncertainty.

CHATGPT
HALLUCINATED NEW
FEATURES
ChatGPT was instructing users to
sign up for Soundslice and import
ASCII tabs to hear audio playback,
but there was just one problem:
Soundslice had never supported
ASCII tab, meaning the AI had
invented the feature out of thin air
while setting false expectations for
new users and making the company
look as if it had misrepresented its
own capabilities.

SEARCH IS STILL
AN UNSOLVED
PROBLEM
User needs are still very
complex and information
gaps persist. Search will
become increasingly agentic
and integrated with other
platforms to solve these
problems.

GOOGLE WANTS TO ANSWER ANY QUESTION YOU HAVE
SEARCH WILL CONTINUE TO PROFOUNDLY CHANGE
Again, we are not going to
revert to 10 blue links as the
primary modality of search. We
need to be prepared for what’s
to come and stop holding to the
past.

NEW RESEARCH
SUGGESTS MOST
SEARCH SIGNALS
ARE NO LONGER
NEEDED
This paper discusses a BlockRank
method using just the query and
documents can perform as well as
using Google’s other signals like
PageRank and Navboost.
Source:
https://arxiv.org/pdf/2510.05396

AI Search Playbook |

Now is the time to shed all the things that
no longer serve us that we inherited in the
SEO industry.
— Me, CEO, iPullRank

AI Search Playbook |

HOW
AI
SEARCH
WORKS
WHAT UNDERPINS OUR STRATEGIC AND TACTICAL FOCUS
92

AI SEARCH IS A CHANGE
FROM DETERMINISTIC TO
PROBABILISTIC
RANKINGS
Google’s launch of AI Overviews
and AI Mode marks a shift from
deterministic rankings—where
content is surfaced largely as-is
based on predefined scoring
signals—to probabilistic rankings
driven by large language models
that interpret, synthesize, and
reason across multiple sources.

BOTH SEARCH AND
GENERATIVE AI ARE BUILT
FROM THE VECTOR SPACE
MODEL
In Search queries and documents
are plotted in multidimensional
space. Documents physically
closer to the query in that space
are more relevant.
In generative AI, the next word is
predicted in that vector space
based on the word relationships
that have been previously seen.

RETRIEVAL-AUGMENTED GENERATION (RAG)
THE CORE PATTERN BEHIND AI SEARCH
At the heart of most AI search platforms is
retrieval-augmented generation. RAG
addresses the fundamental weaknesses of
large language models: hallucinations and
knowledge cutoffs. By grounding
generation in fresh, externally retrieved
data, these systems can deliver answers
that are both fluent and factual.
I introduced the SEO community to this
concept 2 years ago:
https://searchengineland.com/how-searc
h-generative-experience-works-and-wh
y-retrieval-augmented-generation-is-ou
r-future-433393

LEXICAL SEARCH IS DRIVEN BY THE DISTRIBUTION OF WORDS
THIS IS THE FOUNDATION OF SEARCH
The earliest search engines weren’t built to
understand meaning. They were built to
match strings. In the 1960s and 70s,
systems like SMART at Cornell established
the core architecture that would dominate
Information Retrieval (IR) for the next four
decades: the inverted index. If you’ve never
seen one in action, picture the index at the
back of a reference book. Every term has a
list of page numbers where it appears. In
IR, those “pages” are documents, and the
“page numbers” are actually “postings
lists” or ordered references to every
document that contains the term.

SEMANTIC IS
DRIVEN BY A
REPRESENTATIONS
OF MEANING
Instead of treating words as discrete
symbols, you could represent them as
points in a continuous vector space,
where proximity reflected similarity of
meaning. The closer two words were in
this space, the more likely they were to
be used in similar contexts. This leap
from symbolic matching to geometric
reasoning was the conceptual
foundation for embeddings.

HYBRID RETRIEVAL
POWERS RAG
Modern search engines
retrieve passages by
combining lexical and
semantic search then rerank
the documents using
methods like reciprocal rank
fusion (RRF).
Keep in mind 95% of SEO
tools still only perform lexical
analysis.

AI Search Playbook |

TRANSFORMER
GAVE US CONTEXT
The breakthrough came in June
2017, when Vaswani et al. published
Attention is All You Need. This paper
introduced the transformer
architecture, which replaced
recurrence entirely with a
mechanism called self-attention.
Instead of processing one token at a
time, transformers allowed every
token to directly “look at” every
other token in the sequence and
decide which were most relevant for
interpreting its meaning
AI Search Playbook |

THERE ARE VECTOR
REPRESENTATIONS
OF EVERYTHING
Google’s next leap was to embed
everything it cared about in the
search ecosystem. The goal
wasn’t just to improve retrieval. It
was to create a unified semantic
framework where any object, a
website, an author, an entity, a
user profile, could be compared to
any other in the same
high-dimensional space.

ALL CONTENT
FORMATS CAN BE
EMBEDDED IN THE
SAME SPACE
Indexing for AI search is often
multi-modal. Text passages, images,
audio clips, and even tables may be
embedded separately, then linked
under a shared document ID. This
means that an image from your site
could be retrieved directly as
evidence for a generative answer,
even if the text on the page is less
competitive.

GOOGLE INTRODUCED “PERSONAL CONTEXT”

PERSONALIZATION AND MEMORY IN GOOGLE
Google is moving beyond one-size-fits-all
search results by using AI to personalize
responses based on each user’s behavior
over time. This is done through user
embeddings—AI-generated profiles that
represent your interests, preferences, past
searches, and even the types of content
you engage with. These embeddings allow
Google to “remember” patterns and tailor
results in real time, shaping which sources
are selected and how answers are framed.

MUVERA IS MOVING FROM SINGLE VECTORS TO MULTIPLE
LAST YEAR GOOGLE RESEARCH UPGRADED THEIR APPROACH
Multi-vector models, such as ColBERT,
represent each query or document using
multiple embeddings, typically one per
token. They compute relevance via
Chamfer similarity, which measures how
each token in the query aligns with its
closest token in the document. This
method yields more nuanced retrieval
decisions, especially for long-form or
heterogeneous content, but at tremendous
computational cost, especially during
large-scale indexing and retrieval.

HOW GOOGLE’S AI
OVERVIEWS WORK
Google’s AI Overviews no
longer rely on a single search
term to determine what
content to show. Instead, they
expand the user’s query
behind the scenes, pull
relevant insights from
multiple pages, and generate
a synthesized
answer.sources.
AI Search Playbook |

THE AI MODE PROCESS

HOW AI MODE WORKS

AI Mode extends the AI
Overview approach by
pulling in more content
from more variations of
the user’s query—and it
doesn’t just stop at text. It
evaluates passages,
videos, images, and
structured data across the
web, using large language
models to piece together a
comprehensive response.

AI MODE USES
MORE QUERIES AND
SOURCES
AI Overviews aim for brevity and
clarity, so synthesis is
constrained. Think of it as a
single-shot generation pass with
a fixed token budget. AI Mode, by
contrast, is conversational and
persistent. It can run additional
retrieval cycles mid-session,
incorporate follow-up questions,
and adjust the synthesis style on
the fly.

HOW THE QUERY
FAN-OUT
TECHNIQUE WORKS
Google’s AI Overviews and AI Mode
rely on a process called the query
fan-out technique, which expands
the original search into dozens of
related, personalized, and contextual
variations—all behind the scenes.
This allows Google to pull content
from across an entire category,
gathering passages from multiple
sources to generate a more complete
answer.

REASONING IS THE GATEKEEPER OF VISIBILITY

HOW REASONING WORKS IN GOOGLE LLMS
Reasoning is the final stage in
Google’s AI response
process—where the system
decides what to say and how to
say it. After gathering content
from many sources, Google uses
large language models to weigh
the relevance of each passage,
compare them, and stitch
together a cohesive answer.

CRAWL-BASED DISCOVERY
IT’S “REQUESTING” IN REAL-TIME NOT “CRAWLING”
This is the model most content creators
are already familiar with, and how
traditional search engines like Google and
Bing have worked for decades. Crawlers
request a page, extract the links, build a
queue, and keep doing that until they find
the end of the web.
LLMs like ChatGPT and Perplexity don’t
“crawl,” they request, and they don’t index
content or cache the pages so they are
requesting in real time

API-BASED ACCESS
LICENSED OR STRUCTURED FEEDS
This model gives AI systems structured,
direct access to your content via approved
APIs or data partnerships. This is used by
ChatGPT and Microsoft’s Copilot
integrations.
With this model, you retain more control
over what data is shared, how it’s
presented, and how often it’s updated.
However, participation is often limited to
companies with licensing deals or API
infrastructure, and greater control may
come at the cost of reduced organic
visibility.

THE ANSWER GENERATION PROCESS
HERE’S HOW AI SEARCH GENERATES RESPONSES
Answer generation is a five step process
from when the user submits a
query/prompt and system returns a
response.
That process includes expansion, routing,
retrieval, selection, and synthesis.

QUERY/PROMPT IS
A STARTING POINT
IN AI SEARCH
In this new retrieval-and-synthesis
pipeline, the query you type is not
the query the system uses to gather
information. Instead, the initial
input is treated as a high-level
prompt, a clue that sets off a much
broader exploration of related
questions and possible user needs.
The system decomposes the query,
rewrites it in multiple forms,
generates speculative follow-ups,
and routes each variant to different
sources.

AI Search Playbook |

QUERY EXPANSION
AND LATENT INTENT
MINING
The journey from the user’s
initial words to the system’s
full set of retrieval
instructions begins with
query expansion. This is the
stage where the system
broadens the scope of what it
is looking for, aiming to cover
both the explicit and implicit
needs behind the request.

AI Search Playbook |

INTENT CLASSIFICATION
THE QUERY IS BROKEN DOWN INTO NEEDS
This classification step informs everything
that follows, because it sets constraints on
the types of sources and content formats
that will be considered.

SLOT IDENTIFICATION
VARIABLES THAT REPRESENT INFORMATION REQUIREMENTS
Slots are variables the system expects to
fill in order to produce a useful answer.
Some slots are explicit. For example “half
marathon” sets the distance, “beginners”
sets the audience. Others are implicit. The
system may want to know the available
training timeframe, the runner’s current
fitness level, age group, and goal (finish vs.
personal record). These slots may not all be
populated immediately, but knowing they
exist allows the system to search for
content that can fill them.

LATENT INTENT PROJECTION
THE QUERY IS PLACED IN VECTOR SPACE TO IDENTIFY MORE
This is where the original query is
embedded into a high-dimensional vector
space, and the model identifies
neighboring concepts based on proximity.
These are not random; they are informed
by historical query co-occurrence data,
clickstream patterns, and knowledge
graph linkages.

REWRITES AND
DIVERSIFICATIONS
The system then generates
rewrites and diversifications of
the original query. These might
include narrowing variations
(“12-week half marathon plan for
beginners over 40”) or format
variations (“printable beginner
half marathon schedule”). Each
rewrite is designed to maximize
the chance of finding a relevant
content chunk that might not
match the original phrasing.

SPECULATIVE
SUB-QUESTIONS
Finally, the model adds
speculative sub-questions.
These are based on patterns
from similar sessions.
Including these in the
retrieval plan preemptively
allows the system to gather
the material it will likely need
for synthesis.

AI Search Playbook |

SUBQUERY
ROUTING AND
FAN-OUT MAPPING
Once the expansion phase has produced
a portfolio of sub-queries, the system’s
job shifts from what to look for to
where to look for it. This is the routing
stage, and it is where the fan-out map
becomes operational. Each sub-query is
now a small task in its own right, and
the system must decide which source or
sources can best satisfy it, which
modality is most appropriate for the
answer, and which retrieval strategy
will be used to get it.

MAPPING
SUB-QUERIES TO
SOURCES
In routing, the system
maintains an internal mapping
of which source types are most
appropriate for different query
classes. A “plan” might map to
long-form text and structured
schedules; a “checklist” might
map to listicles and product
tables; a “routine” might map
to video; a “definition” might
map to knowledge bases.

AI Search Playbook |

RETRIEVAL
STRATEGIES AND
COST BUDGETING
Routing involves both choosing how to retrieve
information and managing the cost of retrieval.
Retrieval method selection: Some sub-queries
work best with sparse retrieval (e.g. BM25) for
exact term matches, while others benefit from
dense retrieval (embeddings) for semantic
similarity. Hybrid approaches combine both to
leverage their strengths.
Cost-aware budgeting: Each retrieval call uses
resources, so systems allocate retrieval effort
based on sub-query importance—giving
high-priority queries more retrieval passes
from multiple sources, and low-priority ones
fewer or cheaper calls. This is crucial when
using paid APIs or other costly retrieval
sources.

MATCHING THE ROUTING PROFILE
WHERE OPPORTUNITIES ARE WON OR LOST
Content must match the expected
modality or it won’t be retrieved
Ensure multi-modal parity (text,
structured data, transcripts, etc.)
Place content where the routing logic
looks (e.g. API-friendly formats,
transcripts for procedural content)
Align content with routing profiles to
increase retrieval across fan-out
branches

SELECTION FOR
SYNTHESIS
Selection is not only about
relevance to the sub-query. It is
also about the suitability of a
chunk to be lifted, recombined,
and integrated without
introducing factual errors,
formatting issues, or incoherence.
In effect, the system is ranking
not entire pages but atomic units
of information, and the scoring
criteria are tuned to synthesis
needs rather than to
click-through behavior.

AI Search Playbook |

EXTRACTABILITY AS
THE FIRST GATE
If a chunk cannot be cleanly
separated from its
surrounding context without
losing meaning, it is less
valuable to the synthesis
process. This is why content
that is scoped and labeled
clearly tends to survive
selection.

AI Search Playbook |

EVIDENCE DENSITY AND SIGNAL-TO-NOISE RATIO
THE SYSTEM IS LOOKING FOR VERIFIABLE INFORMATION
Once extractability is established, the
system looks at evidence density or the
proportion of meaningful, verifiable
information to total tokens. A dense
paragraph that gives a clear statement,
followed by an immediate citation or
supporting data, is more valuable than a
lengthy, anecdote-heavy section that
buries the facts in storytelling.

SCOPE CLARITY
AND APPLICABILITY
Generative systems are
sensitive to scope because
they are trying to assemble
an answer that is not
misleading. If a chunk does
not make clear the conditions
under which it is true, it is
harder to place it correctly in
the final answer.

AUTHORITY AND
CORROBORATION
The system also weighs the credibility of
the source and the degree to which the
information is corroborated by other
retrieved chunks. Authority in this context
is not limited to domain-level trust; it can
apply at the author or publisher level.
Corroboration is a subtle but important
factor. If three independent, credible
sources agree on a specific mileage
progression, that progression is more likely
to survive selection. Outlier claims may still
make it in if they are well-sourced, but the
system will often prefer information that
has multiple points of agreement.

FRESHNESS AND
STABILITY
Recency is another filter,
especially for topics where
the facts can change. A chunk
that is clearly dated and
shows evidence of recent
review is more attractive to
the model than one with no
temporal markers.

AI Search Playbook |

HARM AND SAFETY
FILTERS
Finally, selection often
applies harm and safety
filters.
These filters can be
domain-specific, drawing on
both explicit policies and
learned patterns from
training data.

WHY GOOD
CONTENT GETS
EXCLUDED
High-quality content can be
excluded from synthesis if it
isn’t easily
extractable—interactive
designs that aren’t crawlable
or long-form narratives that
bury key facts risk being
skipped in favor of denser,
more accessible material.

Engineering for the
Selection Funnel
In practice, this means rethinking
how you structure your content. A
single long page may need to be
designed as a series of clearly
marked, self-contained modules,
each of which could stand alone if
lifted into a generative answer. It
also means pairing each chunk
with whatever metadata, markup,
and alternative formats will make
it easier for a retrieval system to
recognize and use.
AI Search Playbook |

STRATEGIC IMPLICATIONS
AI SEARCH REQUIRES MULTI-DIMENSIONAL APPROACHES
Generative search is not a single ranking
contest for a single query. It is a
multi-stage filtering process in which your
content competes at dozens of points in a
branching, multimodal retrieval plan. The
fan-out means the system is looking for
breadth as well as depth, and the synthesis
step means it is judging your content on
extractability and readiness, not just
relevance.

IT’S JUST SEO,
RIGHT?
NOPE.
133

My data-driven beliefs don’t require you to
agree with me.
— Me, CEO, iPullRank

AI Search Manual |

AI Search uses retrieval so SEO
matters even more than ever.
— You, SEO Person,Some Company

AI Search Manual |

That’s true, but presupposes that the SEO
community has broadly kept up with
advancements in retrieval. The best
practices, software, and common
understanding suggests otherwise.
— Me, CEO, iPullRank
AI Search Manual |

ZIPTIE’S DATA SAYS BEING IN THE TOP 10 GIVES YOU A 25%

CHANCE OF APPEARING IN AI SEARCH
Nobody can run a

business on a 25% chance.
Clearly there must be
something better.

PROFOUND’S DATA SAYS IT’S A 19%

CHANCE OF APPEARING IN AI SEARCH
So, if ranking in Organic
Search doesn’t guarantee
performance in AI Search,
how is it JUST SEO?!

28.3% OF
CHATGPT-CITED
PAGES DON’T RANK
FOR ANYTHING IN
ORGANIC SEARCH!
https://ahrefs.com/blog/chatgpts
-most-cited-pages/

AI Search Playbook |

Show me what you do for GEO that
you don’t do for SEO.
— You, SEO Person,Some Company

AI Search Manual |

It’s deeply disappointing to hear that the
SEO community still struggles with the
distinction between strategy and tactics.
— Me, CEO, iPullRank

AI Search Manual |

There Are Several Different Disciplines That

USE THE EXACT SAME TACTICS

Social Media Marketing Is Just Channel-Specific Content Strategy

Yet, it’s its Own Category

There is no difference between what you
do in Content Strategy and Social Media
Marketing. However, the channel
specificity is big enough that it warrants
its own category.

Doing Things for AI Search Can Also Make Rankings Drop

EACH CHANNEL IS DIFFERENT

You need a bespoke strategy for all
channels. The optimizations that may
work for ChatGPT may cause AIOs to
decline. Your chunking strategy for
Google’s AI surfaces may yield lower
performance in classic search.

CHATGPT USES BING AND GOOGLE AS ITS INDICES
URLS ARE REQUESTED IN REAL TIME THOUGH
Base ChatGPT models do not maintain
their own web index. They are trained on a
massive static corpus, but pull URLs from
indices and request them in real-time.
ChatGPT generates search queries,
sending them to Bing’s API (although new
evidence indicates that they also pull from
Google using SERP API), and retrieves a
short list of URLs. It then fetches the full
content of selected URLs at runtime and
processes them directly for synthesis.

Specific On-Page Factors are

MORE MEANINGFUL TO CHATGPT
Forums with

Pages with

actual

extractable data,

that answer the

relevant

numbers, and

question

discussions

comparisons

Meta
Descriptions

Semantic URLs

Recent publish dates

Source: Profound

BING COPILOT USES ITS INDEX WITH CHATGPT
THE SYSTEM CAN TAKE ACTION IN THE MS365 ECOSYSTEM
CoPilot inherits Microsoft’s full-fledged
Bing ranking infrastructure and then
layers GPT-class synthesis on top. The
consequence is a pipeline where traditional
SEO signals still matter a lot because they
determine which candidates ever make it
to the grounding set, while extractability
and clarity determine whether those
candidates become citations in the final
conversational response.

PERPLEXITY USES BING AND GOOGLE
AND PUTS CITATIONS IN THE FOREGROUND
Perplexity foregrounds its citations.
Sources are displayed prominently, often
before the generated answer itself,
allowing observers to see precisely which
pages informed its synthesis. This
transparency makes it not only a powerful
answer engine for users, but also an
unusually open laboratory for GEO
practitioners seeking to understand what
content earns visibility.

THERE ARE
DIFFERENT
STRATEGIC
CONSIDERATIONS
For example, you like me,
may not want to have pricing
on your site. For SEO that’s
fine, but for generative AI
that means you lose control
over your brand narrative.

I did not do any of that fancy “GEO”
stuff and I’m in ChatGPT in 24 hours.
— You, SEO Person,Some Company

AI Search Manual |

Cool. Now show me that in a competitive
prompt/query space.
— Me, CEO, iPullRank

AI Search Manual |

Chunking is a scam.
— You, SEO Person, Some Company

AI Search Manual |

HERE’S AN ORIGINAL PARAGRAPH
TARGETING [MACHINE LEARNING] AND [DATA PRIVACY]
The development of sophisticated algorithms capable of learning from vast
datasets has revolutionized numerous industries, enabling predictive
models that can identify patterns and make decisions with minimal human
intervention; this process of training models on historical information is the
core of modern artificial intelligence. However, the collection and use of this
data, especially personal information, raise significant concerns about
individual rights and the potential for misuse, leading to the establishment
of regulations like GDPR and CCPA. These legal frameworks mandate that
organizations implement robust security measures, such as encryption and
anonymization, to protect sensitive information from unauthorized access.
The challenge lies in balancing the insatiable need for high-quality training
data, which improves model accuracy and performance, with the ethical
obligation to ensure that an individual's personal details are not
compromised, requiring techniques that can safeguard information while
still allowing for valuable analytical insights to be drawn.

COSINE SIMILARITY
MACHINE LEARNING - 0.6481
DATA PRIVACY - 0.6948

NOW I JUST SPLIT IT INTO TWO PARAGRAPHS
TARGETING [MACHINE LEARNING] AND [DATA PRIVACY]

MACHINE LEARNING - 0.7477
15.4% COSSIM IMPROVEMENT

DATA PRIVACY - 0.7634
9.78% COSSIM IMPROVEMENT

Chunking can improve your relevance.
Anyone telling you otherwise does not know
what they are talking about.
— Me, CEO, iPullRank

AI Search Manual |

You don’t have any case studies that
prove this works.
— You, SEO Person, Some Company

AI Search Manual |

Sure I do. What you’re telling me is that you
don’t though.
— Me, CEO, iPullRank

AI Search Manual |

2X GROWTH IN AI SEARCH VISIBILITY FOR A VEHICLE SALES PLATFORM
Itʼs still early days, but weʼre seeing great results for our clients on AI Search platforms.

661% Growth in ChatGPT,
330% Growth in AI Overviews
Our client in the automotive industry has experienced substantial
growth across all major generative AI search surfaces, including
Google AI Overviews, ChatGPT, and Perplexity. This growth has
been driven by a combination of technical improvements and
strategic content enhancements. Notably, theyʼve achieved
significant visibility gains on non-branded queries that previously
eluded them.

These Are the Result of
Relevance Engineering
Efforts
To drive these results, the client prioritized enhancements
to their JavaScript rendering to ensure full crawlability and
indexation. They also expanded their content footprint with
new, high-intent pages and leveraged detailed competitive
analyses to identify and close semantic gaps. These efforts
were supported by a structured content engineering
process that aligned page copy with user intent and topical
authority.

PAGE

158

HOW A CATEGORY LEADING TELECOMMUNICATIONS BRAND
ACHIEVED 253% GROWTH IN AI OVERVIEWS VISIBILITY
The Results
o 1.41 million impressions in AI Overviews
o AI Overview inclusions grew from 712 to 3,235 in one year
o Steady month-over-month growth in AI-driven visibility

The Background
A major telecommunications provider wanted to expand its footprint in AI Search. Traditional
SEO had secured rankings in organic results, but their visibility in AI Overviews was lagging,
especially for competitive, branded queries where consumers increasingly rely on generative
answers. The brand recognized the need for a strategy tailored to the mechanics of AI search.

What We Did
We applied a content engineering optimization framework, including:
●
●
●

Restructuring content with semantic clarity and answer-like formatting to make it
extractable for AI Overviews.
Enhancing technical accessibility with clean HTML hierarchy, sitemaps, and schema
markup to support generative retrieval.
Creating entity-rich, data-backed passages aligned with how AI models build
summaries.

This combination of technical SEO and AI-native content structuring enabled the clientʼs pages
to be consistently selected for AI Overviews, compounding visibility over time.

GENERATIVE AI CASE STUDY

Our Goals
o Expand visibility in AI Overviews for
branded and competitive queries
o Increase generative search presence and
organic impressions
o Position telecom content as authoritative,
extractable, and AI-ready

Services Used
o Competitive Analysis
o Content Engineering
o Content Audit
o Content Plan

PAGE

159

Until the software catches up, it’s not
realistic to expect everyone is doing the
form of idealistic SEO that would make AI
Search “just SEO.”
— Me, CEO, iPullRank

AI Search Playbook |

WHAT
TO
DO
TIME TO DO SOME RELEVANCE ENGINEERING
161

OUR STRATEGIC APPROACH
FOR DRIVING VISIBILITY IN AI SEARCH

TACTICAL IMPLEMENTATION
5 STEPS TO UNDERSTANDING THE GAPS IN CONTENT
Relevance Engineering is the process of
adjusting content so it’s selected and cited
by Google’s AI systems like AI Overviews
and AI Mode. Unlike traditional SEO, which
focuses on ranking full pages, this
approach targets the individual passages
and concepts that AI uses to construct
responses

EXPAND KEYWORDS WITH QFORIA
THIS HELPS YOU UNDERSTAND YOUR CONTENT GAPS
Qforia extrapolates synthetic
queries based on the initial
prompt and gives you their type
and reasoning similar to what
Google is doing.
https://ipullrank.com/tools/qforia

164

QFORIA HAS BEEN QUIETLY UPDATED
IT DETERMINES THE EXPECTED TYPES OF CONTENT PER TERM
The concept of routing content is
now used as part of the pipeline so
it tells you the type of content
Gemini would consider the best
match for that subquery.

165

QFORIA DATA NOT “REAL ENOUGH” FOR YOU?
REVERSE INTERSECT THE CITATIONS
Pull the AIO citation URLs
and find out what they rank
for using the Semrush API.
Then intersect those terms
to figure out what the “real”
fan-out queries are.
We use FetchSERP as the
SERP API for this.

166

OMNIMEDIA
CONTENT
STRATEGIES
It requires us to think beyond
text and align with the
expected content formats and
locations to drive visibility.

TACTICS THAT WILL RETURN
Cloaking/Double-serving
Brands will serve content as
markdown or add additional
content to their pages for the
LLMs to consume.
Tools like Scrunch AI have
already begun to offer this as a
service.

Persona-driven Content
Brands will develop persona-driven
content to align with the
hyper-personalized user contexts.

Internal Knowledge Graphs
Brands will develop their own knowledge
graphs to scale content production across
multiple surfaces.

Application Integrations
Microsites
Brands will resurrect non-branded
microsites so can generates consensus
of information across the web.

Organizations will build
applications and data pipelines
directly into these surfaces in
partnership with the Google and
OpenAI. The platforms will
default to these unless there is
a clear semantic gap.

Content Partnerships
Brands will engage in content syndication
and development of branded content on
other platforms to curate information
consensus.

168

DOUBLE SERVE AT
THE EDGE WITH
CLOUDFLARE
Cloudflare Workers has a
Markdown Conversion
feature at the edge that allows
you to double serve
markdown to genAI bots.

AI Search Playbook |

MEASUREMENT MUST EVOLVE BEYOND PERFORMANCE METRICS
THREE BUCKETS FOR MEASUREMENT
Signals that show whether your
content and site are aligned
with how AI and search engines
understand and retrieve
information.

Passage Relevance
Entity Salience
Bot Activity
Synthetic Query Rankings

Measures how visible, cited, and
represented your brand is
across search and AI surfaces.

Share of Voice
Citation Rate
Citation Quality
Citation Sentiment

Outcomes that prove whether
visibility and citations translate
into business impact.

Traffic
Events
Conversions
Engagement Depth

TRACK BRANDED
IMPRESSIONS FOR
NON-BRANDED
TERMS
If your brand appears for
non-branded terms that is a
brand awareness impression.
Extract your brand terms from
the AIO response in your SERP
analytics tool and combine the
impression metrics from GSC.

NEW DASHBOARDS
We’ve developed a series of
dashboards that highlight the
AI Search metrics so we have
clarity on the levers that need
to be pulled.

AI Search Playbook |

CONTENT
PERFORMANCE
TRACKING
Monitors passage-level relevance
across multiple AI platforms.
Unlike traditional SEO metrics,
this requires tracking new
performance indicators. Duane
Forrester suggests a “new
dashboard” for the GenAI era,
featuring metrics such as Chunk
Retrieval Frequency, Embedding
Relevance Score, and AI Citation
Count.

AI Search Playbook |

UGC
For queries involving
troubleshooting, product
comparisons, lived experiences,
or niche use cases,
user-generated content (UGC)
and forum discussions are
often prioritized by AI systems.
Generative models value this
type of content because it
reflects authentic, diverse, and
situational insights that can’t
always be found in more
polished corporate content.

ENTITY-RICH,
EMBEDDING-FRIENDL
Y LANGUAGE
Write with clearly defined
entities
Use consistent terminology
Include modifiers and
descriptors: Qualifiers like
size, function, location, and
purpose help differentiate
similar entities.

AI Search Playbook |

STRUCTURED DATA
I’m not here to argue with
you how and when it’s used.
Just know that LLMs can and
do use structured data as part
of RAG pipelines.

ON-PAGE ELEMENTS CHECKLIST

SEO BASICS
❏
❏
❏
❏

Heading hierarchy
Clean, semantic content
Open Robots.txt
Topical Clustering

Yes, a lot of the SEO basic
best practices apply here.
177

WRITING FOR
SYNTHESIS
To ensure your content
performs well in modern
retrieval systems, it’s
essential to structure it in a
way that is both
machine-readable and
human-friendly. Embedding
models rely on clean,
well-defined “chunks” or
semantic units of
information to generate
precise and relevant results.

AI Search Playbook |

RELEVANCE SCORE PASSAGES

USE RELEVANCE DOCTOR
We built a simple tool that scores
passages of content in a layout
aware format. This will improve
your ability to be considered and
extracted.

https://ipullrank.com/tools/relevance-doctor
179

PASSAGE
OPTIMIZATION
Optimizing for extractability
means that content should be
organized into easily defined
sections. Headings and
subheadings should be clear,
and passages should answer
queries directly and
succinctly. The combination
of query/passage is defined as
a semantic unit, and these
units are used to power AI
search.

AI Search Playbook |

Semantic Triples
Semantic triples help search engines
understand context better by
identifying entities, establishing
connections, and building a web of
interconnected concepts, which
provide richer contextual
information beyond just keywords.
These Subject–Predicate–Object
triples are the building blocks of
knowledge graphs, which allow AI
systems to understand relationships
between entities, enabling more
intelligent search results, factual
verification, and structured data for
AI overviews.

USE SEMANTIC
TRIPLES
Semantic triples
(subject-predicate-object)
significantly boost retrieval
accuracy and content
relevance. To do this, always
use the active voice.

PROVIDE UNIQUE,
HIGHLY SPECIFIC,
OR EXCLUSIVE
INSIGHTS
Unique content or proprietary
data increases the likelihood
that your page is retrieved
and cited as authoritative in
RAG pipelines.

AVOID AMBIGUITY
Clearly defined,
straightforward sentences
reduce embedding noise and
retrieval errors.

SIMULATE YOUR RANKINGS BASED ON CHANGES
TOOLS LIKE MARKETBREW ALSO HELP YOU DO THIS
We have an AI Overview simulator that
runs a RAG pipeline and takes updated
content to see what would be returned.
It also makes recommendations on what to
do in order to improve the retrievability.
With so much being open source, there is
no reason that all SEO software companies
cannot offer something similar.
MarketBrew is a company that focuses on
this idea across SEO generally.

TOOLS
FOR
THE
FUTURE
HERE ARE SOME TOOLS I RECOMMEND CHECKING OUT
186

N8N IS A MUST HAVE TO CLOSE THE GAPS
IN THE SEO TOOLS SPACE
We have an AI Overview simulator that
runs a RAG pipeline and takes updated
content to see what would be returned.
It also makes recommendations on what to
do in order to improve the retrievability.
With so much being open source, there is
no reason that all SEO software companies
cannot offer something similar.
MarketBrew is a company that focuses on
this idea across SEO generally.

I prefer open source options so I can
customize them to my needs and I
don’t have to worry about data
privacy.
— Me, CEO, iPullRank
No Code, All Flow |

OLLAMA
It’s an open source platform
and API for running open
source models. I use it for
pretty much everything y’all
use the closed models for
unless I need fidelity with
Google.
https://ollama.com

No Code, All Flow |

Much of what you can do in major chatbots, but Open Source

Ollama Has a Chat Interface

It features a
web search
mode to
perform RAG
and give you
similar
responses to
the major
chatbots.

If you don’t have the model you select, it will go get it for you

It Downloads Models Automatically
It features a
web search
mode as well.

Model size matters too, but you can do most with mid-sized models

It’s Fast If You Have a Decent GPU

It features a
web search
mode as well.

Generative features in crawl data for free

Screaming Frog + Ollama
I generate
embeddings and
qualitatively
assess content,
and as I crawl all
on my own GPU

I RUN N8N OPEN
SOURCE TOO
It’s really easy to get up and
running with n8n locally or
on your own hosted
environment. Just install
NodeJS and run npx n8n.

npx n8n
n8n start
No Code, All Flow |

N8N HAS 1121
INTEGRATIONS
I generate embeddings as I
crawl, asset content, take
screenshots and analyze as I
crawl.
All on my own local GPU.

Use n8n to operate a backend

You Can Build a Full Microservice
You can have the n8n
act as an API that you
can easily integrate
with Google
Workspace or any
other application.

FLEXIBLE HUMAN IN
THE LOOP IS ONE
OF N8N’S BEST
FEATURES
You can set up workflows to
ping you in a variety of places
and only move forward based
on that.

USE THE
TEMPLATES
LOCALLY
The templates available at
https://n8n.io/workflows are
imported into my local
instance from a single click.
Use them as a place to start
with building your own
workflows.

https://n8n.io/workflows/4822-extract-and-analyze-google-ai-overviews-with-llm-for-seo-recommendations/

Automated AIO Analysis
This extracts AIO
details and
generates
recommendations
on how to improve
your content to
appear in AIOs.

AI WORKFLOW
BUILDER FEATURE
MAKES IT EASY
N8N has a chat-driven
workflow builder feature that
allows you to describe the
workflow you want and it
configures it end to end.

AI Search Playbook

IF YOU’RE CONSIDERING AGENTS CHECK OUT CREW.AI
AND BUILD FULLY FUNCTIONAL AGENTS FROM CHAT
Crew is available as a cloud-based
application or open source. You can code
your agents from scratch or describe what
you want in the chat interface and quickly
develop a series of agents to help you scale
you work.
https://www.crewai.com

WRAPPING
UP
CLOSING ARGUMENTS
202

If You Don’t Remember Anything Else…

REMEMBER THESE FIVE THINGS
Search
technology and
behavior has
changed
irrevocably.

It will take
more than SEO
to get you
visibility in the
future

Learn how the
systems work so you
can discover new
opportunities

Most of your

This is an

SEO tools will

opportunity to

not help you

define the future.

get where you
need to go.

204

SCREAMING FROG
+ OLLAMA
I generate embeddings as I
crawl, asset content, take
screenshots and analyze as I
crawl.
All on my own local GPU.

20 CHAPTERS OF
PURE 🔥🔥🔥
Everything you need to know
about how AI Search works.
No vagueries.

ipullrank.com/ai-search-manual

AI Search Playbook |

WE’RE DOING IT EVEN BIGGER NEXT
YEAR HTTPS://SEOWEEK.ORG
https://ipullrank.com

6

THANK YOU | Q&A
Mike King
Founder / CEO
@iPullRank
[email protected]
Tap in with us: ipullrank.com

iPullRank has been proudly featured in:

Award Winning, #GirlDad



本文档为站内渲染。原始文件本地路径：saas/source/seo-llm/raw-seo知识库-TXT整理-seo方法论-ppt-John喜欢和引用的那些海外演讲嘉宾PPTs-FoundConf-dde545.txt（仅本地保留，不入库不部署）