AI SEO is a loose term for improving the chances that your website is discovered, understood and surfaced in AI-powered search experiences as well as traditional search. In practice, much of the useful work is still ordinary SEO: crawlable pages, clear answers, strong information architecture, original content and real authority.
The term sounds newer than much of the work.
That is normal.
Search interfaces are changing.
People now ask questions directly in:
ChatGPT;
Google AI Mode;
Google AI Overviews;
other answer engines.
Businesses naturally want to know how to appear there.
A new industry vocabulary has followed.
AI SEO is not one standard discipline
There is no universally agreed technical definition.
People use "AI SEO" to cover several ideas:
optimising content for generated answers;
making pages easier for AI systems to retrieve;
improving citations in answer engines;
structuring content clearly;
monitoring AI visibility;
and sometimes traditional SEO with a new label.
That makes the term useful as shorthand but dangerous as a promise.
Always ask what the provider is actually doing.
The search interface changed more than the fundamentals
Traditional search often gives the user a list of pages.
AI search may give the user a synthesised answer with supporting citations.
That changes the path.
The website may now contribute information to an answer before the user clicks.
But the AI system still needs source material.
That source material needs to be:
accessible;
relevant;
understandable;
credible enough to use.
Those requirements should sound familiar.
Crawlability still matters
A system cannot retrieve a page it cannot access.
Google needs Googlebot access.
OpenAI documents OAI-SearchBot for ChatGPT Search discovery.
CDNs and security tools can also accidentally block crawlers.
So one part of AI SEO is simply making sure the public information is actually available to the systems you want to find it.
This is not glamorous.
It is essential.
Clear information is easier to use
Suppose a page is titled:
Solutions
and contains four paragraphs of abstract brand language.
A human struggles to understand it.
An AI system has no magical reason to interpret it as a detailed source on commercial plumbing in Longford.
Clear headings and specific language help.
What is the service?
Who is it for?
Where?
How much?
What are the trade-offs?
What happens next?
Good information architecture serves everyone.
Direct answers are useful without becoming robotic
Some people interpret AI optimisation as:
write every paragraph like an FAQ.
I would not.
Natural editorial prose still matters.
But if the article asks:
Does my business need WordPress?
the reader should get an answer near the top.
Then explain the nuance.
That structure is useful because the subject is explicit.
Not because we are trying to write for a robot.
Authority becomes more important when answers are synthesised
If an answer engine is choosing material from multiple sources, generic repetition has weak value.
Why cite a page that merely restates what 1,000 others say?
Real authority can come from:
first-hand experience;
specific examples;
original analysis;
clear professional judgement;
useful local information;
credible references;
strong reputation.
For a small business, that is an opportunity.
You may know your niche better than a giant generic publisher.
Publish that knowledge.
"AI-friendly" content should still be human-friendly
This is the line I would not cross.
I do not want separate text hidden or distorted purely for AI systems.
If a page is useful enough to be cited but awful for the customer who lands on it, the strategy has failed commercially.
The same page should work for both.
That is why good AI SEO is often good web publishing.
What about GEO and AEO?
These are related terms.
GEO usually means Generative Engine Optimisation.
AEO often means Answer Engine Optimisation.
People use them slightly differently.
I cover the terminology separately in SEO vs GEO vs AEO.
The practical lesson is that the overlap is large.
Do not assume three acronyms require three separate websites.
Does schema help AI SEO?
Structured data can help search systems understand page meaning where supported.
Use it accurately.
Article markup.
Business information.
Breadcrumbs.
Relevant supported types.
But neither Google nor OpenAI documents a magical universal "AI citation schema".
Do not mark up claims that are not visible on the page.
And do not assume schema can compensate for weak content.
What about llms.txt?
It is an emerging proposal, not a universal requirement.
Google currently says it does not need new AI-specific text files for its generative Search features.
OpenAI's documented ChatGPT Search guidance focuses on OAI-SearchBot access rather than requiring an llms.txt.
So I would not make that file the centre of an AI SEO strategy.
It deserves discussion.
It does not deserve mythology.
AI SEO can become spam very quickly
Generative tools make content cheap to produce.
That creates the temptation to publish:
every town;
every service;
every question variant;
every comparison;
every possible phrase.
Google has explicitly warned against scaled content created primarily to manipulate rankings or generative responses.
AI does not make doorway pages clever.
It just makes them faster to manufacture.
Useful coverage still needs editorial judgement.
How does this relate to Built by Gavin?
The Guides library is effectively an AI-era SEO strategy without needing to shout "AI SEO" on every page.
Each guide targets a distinct real question.
The article answers it directly.
The deeper content adds nuance and judgement.
Internal links connect the subject graph.
Commercial pages explain the service.
The site remains crawlable and technically clean.
That is the model I trust.
Worth knowing
“AI SEO” is an easy phrase to turn into a new retainer without changing much underneath. I would want a provider to explain the actual work: technical crawlability, content improvement, entity clarity, structured data where appropriate, research, authority building or measurement. A new acronym is not a deliverable.
What would I actually pay for?
If somebody sells AI SEO, ask what the work includes.
Technical crawler access?
Content strategy?
Original research?
Better information architecture?
Search Console work?
Entity/business information?
AI referral measurement?
Those can be real services.
If the answer is:
We add AI keywords and submit you to all the models,
I would want a much more precise explanation.
Gavin's take
I would be very wary of anyone selling guaranteed AI citations. The useful goal is to make the business a strong, accessible and genuinely useful source. Whether a particular answer engine cites a particular page on a particular query is not something a web designer can honestly promise.
Can AI SEO guarantee citations?
No.
OpenAI says ChatGPT Search placement is not guaranteed.
Google says crawling, indexing and serving are not guaranteed.
A provider can improve eligibility and quality.
They cannot promise that proprietary systems will cite you for a particular question.
The difference between influence and control matters.
The bottom line
AI SEO is best understood as the extension of search optimisation into AI-powered discovery.
The useful work includes:
technical accessibility;
clear answers;
original expertise;
strong internal structure;
accurate business information;
authority;
and measurement.
The interface changed.
The need to publish something worth finding did not.
Want your website prepared for both traditional and AI-powered discovery?
See my website approach or tell me what subjects you want the business to become known for.
I would rather build durable search visibility than sell a temporary acronym.
---