AI Search Optimisation
Being found when the answer is written for you
AI Overviews, ChatGPT, Copilot, Gemini and Perplexity increasingly answer the question instead of listing ten links. We work on what genuinely influences whether your business is part of that answer — and we are straight about what nobody can control.
What actually changed
The click is no longer the only outcome
For twenty years the job was to be one of the ten links. Increasingly, the result is a written answer, and your business is either part of it or absent from it. That does not make search less valuable — people still buy things — but it changes which pages earn the visit.
The pattern we see in the data is consistent: impressions hold up, clicks fall on informational queries, and commercial queries are largely unaffected. In other words, the top of the funnel gets thinner and the bottom does not. That has strategic consequences, and most of them argue for investing more in the pages closest to a purchase, not fewer.
The second change is subtler. These systems assemble a picture of your business from many sources at once — your site, directories, coverage, reviews, social profiles, other people’s descriptions of you. If those sources disagree, the picture is muddled, and a muddled entity is easy to leave out of an answer.
The inputs that genuinely matter
Four things that make a business usable in an answer
None of these are new tactics. What is new is how much they now decide, and how badly they are usually executed.
Access
The content has to be reachable and readable. That means crawlable pages, meaningful HTML rather than client-only rendering, and a deliberate decision about which AI crawlers you allow. Sites lose here before anything else is even considered.
Clarity
A direct answer near the top of the page, headings that describe rather than tease, definitions stated plainly, and figures that carry their source. Writing that survives being lifted out of context and quoted on its own.
Consistency
Your name, what you do, where you operate, who you serve — described the same way everywhere a machine could look. This is unglamorous cleanup work and it moves the needle more than any markup.
Corroboration
Independent sources saying the same thing about you. Coverage, mentions, citations, real reviews. This is the slowest input, the hardest to fake, and consequently the one that carries the most weight.
Our position, stated plainly
There is no markup, no file and no prompt that makes a language model cite you. There is only being reachable, being clear, being corroborated, and being worth quoting.
We will happily be the agency that talks you out of an AI search retainer. If your fundamentals are not in place, this work has nothing to stand on, and the responsible sequence is to fix that first.
Where it does make sense — usually for businesses in considered, research-heavy purchases — the work is real, specific and worth doing. It is just not magic, and we will not describe it as though it is.
Scope
What the engagement actually involves
- A baseline: how the major assistants currently describe your business, your category and your competitors, against a fixed question set
- Crawl and access review specifically for AI crawlers, plus a documented decision on which to allow
- Answer-readiness work on the pages that matter — direct answers, descriptive headings, sourced figures, clean structure
- Entity consistency cleanup across your site, profiles, directories and third-party descriptions
- Structured data reviewed for validity and for agreement with what the page visibly says
- Original data or methodology, where your business is in a position to publish something genuinely new
- Digital PR aimed at corroboration in sources these systems already trust
- Quarterly re-measurement against the same question set, with an honest read on movement
Honesty
What we will not sell you
What is real
- Making your content reachable, readable and quotable
- Fixing how inconsistently your business is described across the web
- Publishing information that is genuinely worth citing
- Earning corroboration from sources that carry weight
- Measuring representation over time and reporting it honestly
What is not
- "AI schema" — it does not exist
- A file that makes language models cite you
- Guaranteed inclusion in AI Overviews or any assistant
- Citation-share statistics we cannot source
- Prompt tricks, hidden text, or any attempt to manipulate a model
Questions
The questions worth asking any AI search agency
Is AEO just SEO with a new name?
Largely, yes — and anyone telling you otherwise is selling something. The inputs that make a page eligible to be used in an AI answer are the inputs that already made it a good search result: it can be crawled, it is clear, it is accurate, it is corroborated elsewhere, and it says something worth repeating.
Two things are genuinely different. First, writing has to survive extraction: a claim pulled out of your page and dropped into an answer has to still be true and still make sense without the paragraph around it. Second, consistency of how your business is described across the open web matters more than it used to, because these systems assemble a picture from many sources rather than ranking one.
Can you guarantee we will appear in AI Overviews or be cited by ChatGPT?
No. Nobody can, and a guarantee is a reliable signal that you are talking to the wrong agency.
These systems are non-deterministic, they change without notice, and the same question can produce different answers to different people on the same day. We work on the inputs that make you eligible and easy to use. We do not sell an outcome we do not control.
Does adding special schema or an llms.txt file make a difference?
There is no "AI schema", and no file you can place on your server that causes a language model to cite you. Structured data helps machines understand a page, and it should be valid and match what the page visibly says — that is the whole of the honest claim.
If an agency’s AI search offering is mostly a markup checklist and a text file, you are being sold the appearance of a service.
How would we measure any of this?
Imperfectly, and we will say so up front. There is no equivalent of rank tracking here, and traffic attribution from assistants is patchy.
What we do measure: how your brand and category are described across the major assistants over time, against a fixed set of questions your buyers actually ask; where you are named and where a competitor is instead; impression-to-click patterns on informational queries in Search Console; and referral traffic from assistants where it is identifiable. It is a directional picture, and we present it as one.
Should we block AI crawlers?
It depends on what you are protecting and what you want. Blocking reduces the chance of your content being used without attribution, and it also reduces the chance of being represented at all.
For most businesses selling a service, being findable is worth more than being unquoted. For publishers whose product is the content itself, the calculation is genuinely different. We will walk through the trade-off for your situation rather than applying a default.
Find out how AI systems currently describe your business
We will run your category and your company through the major assistants before the call, and show you what comes back. Sometimes it is fine. Sometimes it is wrong in a way worth fixing immediately.
Related
Where to go next
- answer engine optimisationMaking a page usable as an extracted answer.
- generative engine optimisationBeing one of the sources a generated answer draws on.
- Google AI OverviewsDiagnosing what answer surfaces are doing to your clicks.
- ChatGPT visibilityWhat assistants say when asked about your category.
- organic searchThe foundation this work depends on entirely.
- our standardsIncluding the claims we refuse to make.
- brand and messagingHow consistently you are described matters.
Last updated · Reviewed by Zubair Afzal