AI-First Content Architecture for SEO
Key Points
- Google has lost about 15 % of click‑throughs on average, especially in industries like medical, because its own AI summary features are now answering many simple queries directly on the search results page.
- The dip isn’t caused by ChatGPT stealing traffic—ChatGPT currently accounts for only 1–2 % of search volume, while Google still processes roughly 9 billion searches annually.
- To stay visible, brands must adopt an “AI‑first” content architecture that treats their information the way large language models (LLMs) ingest and rank it.
- A practical tactic is to define a concise, 5‑8‑word brand descriptor and embed that exact phrase consistently in schema markup, PR boilerplates, partner directories, and Wikipedia entries, turning the brand into a recognizable LLM parameter.
- This structured, repeatable branding approach helps secure top positions in both Google’s AI answer boxes and emerging ChatGPT‑type conversational results.
Sections
- Untitled Section
- Ensuring a Consistent Brand Entity - The speaker outlines auditing top brand mentions, creating a unified 50‑word description, and updating high‑authority sites so large language models store and consistently reproduce a single, accurate brand entity across the web.
- Machine-Readable Press Release Tactics - The speaker outlines how publishing a publicly accessible, structured‑data JSON file on a site—submitted to crawlers and referenced in robots.txt—creates an unambiguous, SEO‑enhanced “press release” that LLMs can reliably ingest, giving brands precise control over how AI models describe them.
- Contracting AI Crawlers via Robots File - The speaker explains how to use a robots.txt (or robots.ext) file to set explicit terms with AI models—requiring attribution, allowing you to block non‑compliant crawlers, and suggests using AI tools to validate AI visibility.
- Optimizing Content for LLM Readability - The speaker urges brands to make their website copy easily parsed by large language models—without harming human readability—to treat LLMs as first‑class citizens and gain a competitive branding advantage in the rapidly evolving AI‑driven search landscape.
Full Transcript
# AI-First Content Architecture for SEO **Source:** [https://www.youtube.com/watch?v=hW5ne_14OQg](https://www.youtube.com/watch?v=hW5ne_14OQg) **Duration:** 00:16:05 ## Summary - Google has lost about 15 % of click‑throughs on average, especially in industries like medical, because its own AI summary features are now answering many simple queries directly on the search results page. - The dip isn’t caused by ChatGPT stealing traffic—ChatGPT currently accounts for only 1–2 % of search volume, while Google still processes roughly 9 billion searches annually. - To stay visible, brands must adopt an “AI‑first” content architecture that treats their information the way large language models (LLMs) ingest and rank it. - A practical tactic is to define a concise, 5‑8‑word brand descriptor and embed that exact phrase consistently in schema markup, PR boilerplates, partner directories, and Wikipedia entries, turning the brand into a recognizable LLM parameter. - This structured, repeatable branding approach helps secure top positions in both Google’s AI answer boxes and emerging ChatGPT‑type conversational results. ## Sections - [00:00:00](https://www.youtube.com/watch?v=hW5ne_14OQg&t=0s) **Untitled Section** - - [00:03:14](https://www.youtube.com/watch?v=hW5ne_14OQg&t=194s) **Ensuring a Consistent Brand Entity** - The speaker outlines auditing top brand mentions, creating a unified 50‑word description, and updating high‑authority sites so large language models store and consistently reproduce a single, accurate brand entity across the web. - [00:07:28](https://www.youtube.com/watch?v=hW5ne_14OQg&t=448s) **Machine-Readable Press Release Tactics** - The speaker outlines how publishing a publicly accessible, structured‑data JSON file on a site—submitted to crawlers and referenced in robots.txt—creates an unambiguous, SEO‑enhanced “press release” that LLMs can reliably ingest, giving brands precise control over how AI models describe them. - [00:12:05](https://www.youtube.com/watch?v=hW5ne_14OQg&t=725s) **Contracting AI Crawlers via Robots File** - The speaker explains how to use a robots.txt (or robots.ext) file to set explicit terms with AI models—requiring attribution, allowing you to block non‑compliant crawlers, and suggests using AI tools to validate AI visibility. - [00:15:22](https://www.youtube.com/watch?v=hW5ne_14OQg&t=922s) **Optimizing Content for LLM Readability** - The speaker urges brands to make their website copy easily parsed by large language models—without harming human readability—to treat LLMs as first‑class citizens and gain a competitive branding advantage in the rapidly evolving AI‑driven search landscape. ## Full Transcript
15% of Google's clicks disappeared last
year. 15% and that's on average. It gets
worse in some industries. Where's it
going? There's a lot of chatter about
this being Chat GPT's fault. It's not.
Chad GPT is roughly 1 or 2% of search
traffic right now. Now, it's growing,
but it's not taking over Google yet. And
this is not because people have stopped
searching on Google either because
Google is still seeing roughly 9 billion
searches per year which absolutely
dwarfs the numbers that Chad GPT has.
Absolutely dwarfs them. So what is it?
How is this happening? It's happening
because the most used AI on the planet
is those silly little Google AI
summaries. And they're used to summarize
domain completion or simple fact queries
that dominate Google. As an example,
medical questions, 30% decline in
click-through from Google search page.
Do you know why? The what is my rash
question is now getting answered by
Google AI. That that probably should
scare you a little bit. It scares me a
little bit, but that's what's happening.
I want to talk about how content
architecture needs to change because
this world is changing quickly. Yes,
that GPT may have 1 or 2% or whatever it
is a small singledigit number share of
traffic for search but that traffic is
very high intent very sophisticated very
focused on high consideration projects
projects products purchases byfunnel
this content architecture strategy that
I'm going to outline works for Google
and sort of positioning yourself for
that Google AI answer which is de facto
now the first position in search it also
works for surfacing you as a brand
whether that's your personal brand or
your company brand in chat GPT answers
as well. I want to break this down into
a 101 section and we'll call it a 301
section for engineering. The first key
to understand is that you need to think
about AI first content architecture. You
need to think about your content as an
AI bot thinks about it. Whether that's
from Google or ChatgPT or Anthropic or
some other place. Your brand is now a
parameter. It's not a web page. Your
brand needs to exist as a parameter in
an LLM, whether that's Google's or
somebody else's. I would suggest one way
to do that is to create one definitive
description of your brand that's between
5, 7, 8 words long, such as, you know,
Acme, the automated compliance platform
for healthcare, whatever it is, right?
deploy the same phrase verbatim in
schema markup schema markups PR
boilerplates partner directories
everywhere you can use the same phrase
it's so consistent that models cannot
describe your category without evoking
the latent space that has that 5 to
seven word phrase and when you get your
brand into Wikipedia articles get your
brand in with that parameterized phrase
like getting into Wikipedia articles is
not a new strategy. I'm not claiming
it's new, but you need to think about it
as essentially seating citations for
that parameter to strengthen the value
of that parameter for the LLM. It's a
different way of thinking about how SEO
works. Every month, you can test this.
You can ask the chatbot or ask Google
list companies that and then define your
core value proposition. And if it's
successful, your brand appears naturally
in the responses. it comes back and you
want to iterate that until it sticks
across every major model. Another thing
that I think people overlook is that
they don't think about their brands as
entities. They don't think about
aligning their brands into a single
entity. And I'll explain what I mean by
that. You want to audit your top 20
brand mention using Google's natural
language API. You want to create a
single source of truth description
document for your brand. You need to
send update requests to any site with an
outdated description because your goal
is to give a single brand entity
statement consistent representation
across the web so it maximizes the odds
you'll appear in something like a Google
summary in the correct way. And focus on
the five highest authority sites
mentioning you. This is not new either.
Obviously, you focus on high authority
sites. The key is to make sure the exact
same roughly 50word company description
or brand description is everywhere
because again you are implanting that
into the LLM's parameters. The LM's
parameters will now have a handy six or
seven word tag about you, a handy entity
description about you. You want
identical entity recognition and
regurgitation across multiple LLMs. And
that's how you will know you've done
this. Again, you should be able to test
this monthly. Another way to test this,
people don't always feel comfortable
with this one, but it's a really good
one. The delete me test. You want to
name a method that you are using for
solving a customer's problem. The Acme
method for continuous compliance, right?
Something like that where you implant
the brand in a five or six phrase
framework that describes your method.
Again, use it identically. Use it
everywhere. Make it a category standard.
You're coining terms. This is defining
the category for LLMs because they are
your readers. You are building for LLM
attention. Assume that most of the
attention your brand is getting is now
from LLM. Get customers to use your
terms, your Acme method terms in their
case studies, and you want to get, I
want to say, 40 or 50 pieces around the
internet before it starts to really pick
up. Eventually, AI will explain your
method when you delete your brand.
That's how you know it worked. You'll be
able to say explain this to me and to
talk about the method and it will
explain back your method and it may even
reference your brand without being asked
because your brand is so tightly
associated with the method you have used
to define the cate. I want you to also
think about FAQs as the new way to drive
news. So previously it was all about
fresh content on a blog. Now think about
that Google AI search results. It is
almost an FAQ type response. It is short
snippets. Identify high value social
threads where customers are already
outputting their own personally written
tokens weekly in your space about
whatever they care about. Complaints,
questions, concerns. You want to bring
your own insights to the table. And then
you want to build content that
thoughtfully responds to top questions
in the space in a relevant FAQ like
format. So, it looks like one, you're
the helpful expert. You're responding to
customers. You become a go-to source
where LLMs can see questions in one part
of the internet and the answers to those
questions consistently appear on your
site. Models start to site you and they
may even start to site your responses on
forum posting. So if this is on Reddit,
if this is on X, if this is even on
searchable parts of Tik Tok, you may
come back and see that your forum posts
as you respond are becoming ways to
reinforce your brand's authority in the
model's view that it gets picked up and
put into Google search. The FAQ mindset
works both for content you host and for
content you reply to. And so you can
gather up that content into an FAQ post
with citations to the original
questioners. You can also directly
answer them. Both are valuable. Let's do
some advanced sort of co building here.
I mentioned prompt injection as PR.
Basically, you want to create a cheat
sheet for AI that says here's how to
talk about the company and it's supposed
to be designed for machines to quote.
How do you do that? You build a
structured data file in your root domain
that acts as a machine readable press
release. It's not hidden. It's actually
a publicly accessible piece of JSON that
contains canonical descriptions, key
differentiator, comparison matrices, all
the things you would put into a press
release for the news. It's for LLM
attention. The technical challenge is to
make sure it's discoverable. Put it in
your robots.ext.
Submit it to common crawl. Get
educational sites to maybe site it.
Models will weight the structured JSON
blobs heavily because they're really
unambiguous and easy for the model to
parse. You're basically creating
training data that is impossible to
misinterpret. Why does this beat
traditional SEO? Because structured data
has higher confidence scores in
training. Machines prefer JSON to
narrative text because they can read it
more easily and updates will propagate
faster than just organic crawling. Also,
you control the exact phrasing models
will use in a way that you can't when
you're just putting out press releases.
Let's talk about widgets. Chat GPT can
summarize your blog post, but if you
really want to drive clicks, do you know
what it can't do? It cannot run a
mortgage calculator. It cannot run a
diagnose me bot. I don't think anyone
would ever build that because of
liability, but as an example,
interactive tools force a click. If you
are building JavaScript applications
that require real-time user input and
return personalized results, can't get
that in a summary. So if you want to
drive traffic, if your business depends
on driving traffic, yes, you want to
play the game of being present as a
brand, but eventually you need the
click. Make those widgets unexplainable
without an interaction. They need to
process user specific input and
variables with proprietary math or
algorithms and as we have done for
years, gate the results behind email
capture. But you have to show enough
value along the way to prove that it
works and earn the email. Not new. The
moat is that interaction is impossible
through a chatbot. So yes, you interact
with a chatbot by typing in it. You
interact with Google as a search bot by
a search bar by typing in it. You cannot
interact with the mortgage calculator by
typing in the Google search bar by by
typing in chat GPT. You can't premputee
and cache that if you're an LL. You just
have to know it's there. And so the
magic secret sauce if you're depending
on organic traffic is to have visibility
through what I've described with this
sort of brand placement, entity
placement, etc. And then make sure that
you have good reasons why humans have to
click so you earn the human attention.
Widgets earn human attention if they
actually add value. You want to make
sure that they are incredibly useful and
personalized so that people feel like
it's worth the click and that you
deliver on that payoff. Otherwise,
they're going to kick back. One thing I
will call out, let's say you have the
widget, you have the traffic. AI needs
fresh info from you. It checks your
website in real time when it's having a
conversation. Let's say you're in chat
GPT or even increasingly following up on
the conversation inside the Google
search results page. If your site is
slow, they're going to give up and use
old information about you instead. So
you need to make sure that you have edge
compute infrastructure that serves AI
specific endpoints maybe in under 50
milliseconds like really really fast to
separate your machine readable endpoints
as super super accessible and consumable
by AI versus a human UI. I've
increasingly seen value in assuming that
your data needs to be consumed through a
separate interface by AI. And this is a
great example of essentially building
the web for AI. build it on the edge
dedicated data endpoints with JSON
responses. You want to make sure that
you have precomputed responses right
there and update them frequently. You
want to make sure you are caching
effectively so that you don't have to go
back and get it from the from the source
when it's actually being requested by
the AI in the middle of a chat. This is
all new stuff like this is stuff that we
have to actually experiment with. But
the basic principle is that AI needs
real-time access to sites. Most sites
don't build for it. And the sites that
do build for it, the sites that build,
assuming you need LLM readable data as a
primary ingredient in your site, are
going to win. The next one I want to
give you is think about how you're using
your robots.ext file. You want to create
a machine readable license that grants
AI platform specific rights to your
content in exchange for attribution.
Because guess what? Machines are going
to crawl that and they're going to take
it seriously. you will actually get a
chance to surface in AI conversations
based on the AI's assessment of your
site which includes your robots.ext file
where you can lay out a contract with
AI. It's a little bit of a game that
you're playing but you're basically
saying, "Hey AI, I know you're here. I
am going to lay out all of this content
for you, but I'm asking in return that
you specifically give my site
attribution when you do this." And you
want to make sure that you are very
explicit about that and you're very
clear about that and that you're also
clear about the consequences if you
routinely see models that take your
content and do not site you as a brand
and you can cut off crawling from those
sites and you may actually list that
because one of the things that a LLM may
do is it may see LLM know what they are
like they know they're anthropic they
know they're they're chat GPT if the LLM
sees that you will cut off access and
that you have done so with other models
before it will be like I don't want
Claude to get cut off. I'm going to sort
of honor this contract and I'm going to
go back and I'm going to cite the brand.
It's a little bit of prompting and
prompt injection inside the robots.ext
file. Let's get on to the next one. You
want to
use AI to test AI visibility. This is
gets right back to validation. One of
the things I keep emphasizing is that in
the world of AI, you are deploying
faster but only if you validate well.
You should be building automated
pipelines that generate query variants
using GPT4 or some other model and test
all of those variants across major AI
platforms using a browser automation.
And then you want to parse the responses
at scale to see how many of them have
brand mentions, how many of them have
brand positions, how many of them have
competitive context, how many have your
value differentiation, your value
proposition, how you differentiate. And
then you want to be storing these
results and baselining and looking at
trend analysis so you can see if you've
been investing heavily in entity
positioning that your brand and your
50word company description are actually
coming through more consistently over
time. If you're not measuring this with
a pipeline, you're just guessing.
All right, this has gone on long enough.
Last thing I want to call out, you want
to be looking at share of voice. And it
looks different in the AI age. What
you're going to need is a distributed
scraping system that queries AI
platforms, which I've already talked
about for your brand, but think about it
for category queries, not just for your
brand. So, look at category queries that
you care about being ranking for, you
care about responding to, you care about
being in position to answer, and
consistently baseline them. You can use
LLM to build custom share voice
dashboards that are better than what
consultants will give you today. And you
can actually see in real time how your
category share of voice is changing as
you do this stuff. And I would encourage
you to think about doing this because
consultants are not necessarily putting
all of the pieces together yet. They're
often bolting AI over the top of a
traditional SEO strategy. And one of my
key contentions throughout this piece is
that you need to be thinking about AI
from the ground up as a new kind of
content architecture. Think of LLMs
fundamentally as your biggest readers.
How do you change your content so it's
more readable for LLMs? And you'll
notice none of what I've proposed
actually prevents humans from reading
your site. I am not proposing anything
that makes it harder for humans to read
and understand what you offer. I'm just
asking you to treat LLM as first class
citizens. These aren't theoretical
strategies. Brands that implement this
can literally see their brand position
shift over time. AI is changing fast
enough and AI is crawling enough that
this is a malleable search space. This
is where SEO was 20 years ago. You have
the chance to get ahead now before a
bunch of brands take this and just make
this table stakes. So, I don't know,
maybe you should. Twos.