AI search optimization
Ask an AI who to call in your city. Somebody gets named. Is it you?
Answer engines do not list ten links. They assemble one answer and cite a handful of sources. Being in it depends on whether your content can be lifted, and whether anything else on the web agrees with you.
Measured monthly against real customer questions · No guarantees sold, because none can honestly be made
What an answer engine needs from you
- A liftable answerForty to sixty words that stand on their ownContent
- One entityThe same name, address and phone everywhereIdentity
- CorroborationSomeone other than you saying itTrust
- Structured dataSo the machine does not have to guessPlumbing
What is AI search optimization?
AI search optimization is the work of getting a business named accurately when someone asks an AI assistant a question it could answer. It requires three things classic SEO does not fully cover: answers that stand alone when lifted out of a page, business details consistent everywhere, and third-party sources that corroborate the claims.
The honest framing matters here, because this field is full of confident claims built on very little. There is no primary documentation from OpenAI, Google, Anthropic or Perplexity explaining how sources get selected. What exists is vendor research of varying quality, plus a set of things that are obviously true: an engine cannot cite a page it cannot crawl, cannot quote a paragraph that needs the one above it for context, and will not confidently describe a business whose own details contradict each other.
So the work here is not a growth hack. It is making the business unambiguous and quotable, then measuring whether it shows up. That is unglamorous, and it is the only version of this that can be delivered honestly.
Several Burlington roofers are well reviewed, with differences in warranty length, whether they self-perform, and how quickly they can get on a roof after a storm.
Your businessDirectoryReview site
Four channels, not one
Each engine reads a different index
| Engine | Retrieval backend | What that means for you |
|---|---|---|
| ChatGPT | Bing index plus training data | Bing Webmaster Tools health matters more than most local SEO plans acknowledge. If you are not in Bing, you are largely invisible here. |
| Google AI Overviews & AI Mode | Google index, with query fan-out | Reported overlap with the classic top ten has fallen sharply. Ranking first no longer implies being cited. |
| Perplexity | Its own pipeline over several search APIs | The closest to classic rankings of the major engines, and the one that cites the most sources per answer. |
| Claude | Brave Search | A different index again. Almost nobody optimises for it, and almost nobody checks whether they appear in it. |
Reported overlap between which domains each engine cites is low. Treating "AI visibility" as one channel is the most common planning mistake, and it is why Bing Webmaster Tools — which almost no local SEO engagement touches — is a first-class concern here.
What's included
The work, itemised
An extraction audit of your content
Which pages contain an answer an engine could lift whole, and which are prose that only makes sense from the paragraph above. That distinction decides almost everything here.
Question-shaped restructuring
Headings written as the question someone actually asks, followed immediately by a 40–60 word self-contained answer, then the elaboration. Applied to every page that matters commercially.
One connected entity graph
Organization, service, breadcrumb and person structured data linked by stable identifiers, so an engine resolves your business to one entity rather than three similar ones.
Byte-identical business details
Name, address and phone matched exactly across the site, Google Business Profile and every directory. Inconsistency is the most common reason an engine cannot confidently say who you are.
A facts page worth citing
Copy-pasteable boilerplate, founder bio, verified numbers and logo, so anyone writing about you — human or model — gets it right without inventing anything.
Bing and IndexNow health
Bing Webmaster Tools verified, sitemap submitted, IndexNow wired up. Skipped by almost every local SEO engagement, and it is the retrieval layer behind ChatGPT.
Crawler policy that lets you be found
A robots.txt that admits the answer bots — OAI-SearchBot, Claude-SearchBot, PerplexityBot — and does not block Google-Extended, which governs Gemini grounding and is not a ranking signal.
Third-party presence
Directory profiles, association listings and the local roundups engines lean on. Most citation-driven visibility reportedly comes from sources other than your own homepage.
How it works
Ask, fix, restructure, re-ask
Ask the engines about you
Week 1We run the questions your customers would ask across all four engines and record what comes back — whether you appear, who does instead, and whether what is said about you is even accurate.
Fix what the engines cannot verify
Weeks 2–3Inconsistent business details, missing or broken structured data, and claims on your site that nothing else on the web corroborates. Verifiability comes before persuasion here.
Restructure the commercial pages
Weeks 3–8Question-shaped headings, self-contained answers, comparison tables where a decision is involved. Tables and lists are reportedly cited far more than unstructured prose.
Re-run the questions, and keep re-running them
MonthlyThe same prompts, monthly, recorded. This is the only honest measurement available — Search Console reports AI impressions but no clicks, and the other engines report nothing at all.
How do you make content extractable by an AI answer engine?
Write the heading as the question someone actually asks. Follow it immediately with a 40 to 60 word answer that makes complete sense on its own, with no pronouns pointing at earlier paragraphs. Then elaborate underneath. If the answer needs the previous sentence to be understood, it cannot be lifted.
This page is built that way, and so is every page on this site. The pattern is enforced by a build-time check: an answer block outside the word range fails the build rather than shipping. That is not a flourish — it is the only way a rule like this survives the third month of a busy content schedule.
The second half is harder and slower. An engine weighs what other sources say about you far more heavily than what you say about yourself. Directory profiles, association listings, local roundups and genuine reviews are what turn "this business exists" into "this business is a reasonable answer to that question", and they accumulate over months.
Fit
Who this suits, and who it doesn't
This is a good fit if
- Your customers ask comparison questions — which company, what does it cost, who is good in this area.
- You already have or are building solid classic SEO. This layers on top; it does not substitute.
- You can accept measurement in impressions and appearance logs rather than clicks.
- You have real, verifiable specifics about your business worth citing.
- You want to be early in a channel where almost nobody local is doing deliberate work yet.
This isn't the right fit if
- You want a guarantee of being recommended by ChatGPT. Nobody can honestly sell that.
- You have no website content worth structuring. Fix that first — this is a structuring discipline.
- You need measurable leads this quarter. Paid search is the honest answer to that brief.
- You want AI-generated pages at scale to blanket the results. That is scaled content abuse, and we will not build it.
Related
What this works with
SEO & AI search
Rank for the jobs you want in Google, and get named when someone asks ChatGPT or an AI Overview who to call.
Website design & build
Fast sites built to turn a click into a phone call, with the tracking wired in before launch, not after.
Lead response & follow-up
Enquiries answered in minutes at any hour, quotes chased for a month, reviews requested automatically. Email sequences included.
Further reading: how AI is reshaping search, and what a business can do about it.
Questions about AI search
What is AI search optimization?
It is the work of making a business findable and quotable by AI answer engines — ChatGPT, Google AI Overviews, Perplexity and Claude. It overlaps heavily with SEO but adds three specific demands: self-contained extractable answers, consistent entity data, and third-party corroboration.
How is it different from SEO?
SEO competes for a position on a page of results. AI search competes to be included in an answer that is assembled and cited. The foundations are shared — crawlable, accurate, well-structured content — but the content shape and the measurement are genuinely different.
Can you guarantee ChatGPT will recommend my business?
No, and be careful with anyone who says otherwise. There is no primary documentation from OpenAI, Google or Perplexity describing how citations are selected. What can be done is making a business easy to find, easy to verify and easy to quote — then measuring whether it appears.
How do you measure AI visibility?
Two ways, both imperfect. Google's Search Console generative AI report gives impressions across AI Overviews and AI Mode — but no click data at all. For the others, we run a fixed set of your customers' real questions monthly and record what comes back. That log is the honest measure.
Should I publish an llms.txt file?
You can, and expect nothing from it. Server-log studies across hundreds of domains recorded no requests for it from the major AI crawlers, and Google has said Search ignores it. We generate one automatically because it costs nothing. Anyone selling llms.txt as a strategy is selling 2024 advice.
Does blocking AI crawlers protect my content?
It removes you from the results. Search bots and training bots are separate controls, so you can allow one and block the other — but blocking OAI-SearchBot or Claude-SearchBot means not appearing in those products at all. For a small consultancy or contractor, that trade rarely makes sense.
Is FAQ schema still worth adding?
For rich results, no — Google retired FAQ rich results in May 2026. The markup is harmless, may help machines parse the page, and costs nothing to ship. Any 2026 proposal that budgets real effort for "FAQ rich snippets" is working from outdated assumptions.
How long before this shows results?
Factual queries — who you are, what you do, where you work — can improve within weeks once the entity data is consistent. Recommendation queries take longer, because they depend on third-party sources agreeing with you, and those accumulate slowly.
Find out what the engines currently say about you.
We will run your customers' real questions across ChatGPT, Perplexity, Claude and Google's AI results and send you what comes back. Sometimes the finding is that you are absent. Sometimes it is worse — you are described wrongly.
You talk to Mathew.Not an account manager, and not a sales team.