AI search visibility
Your customers have started asking an assistant instead of browsing ten results. We make your website one of the sources that answer is built from — and we tell you honestly how far that has got.
Search used to hand people a list of places to look. AI search hands them an answer — and names the sources it trusted. If you are not named, you are not in the conversation.
In a study of Google AI Overview citations, 55% came from the top 30% of the cited page. Where your answer sits on the page is not a detail — it is most of the job.
What changed
This is not a prediction about the future. It is what is already happening to the enquiries that used to arrive through ordinary search.
A customer used to type a question, scan ten blue links and pick one. Now an assistant answers in a paragraph — and often that paragraph is the whole interaction. If your business is not inside it, you were never in the running.
Generative engines retrieve a handful of pages and write from those. You are competing to be one of the retrieved few, not to hold a position on a list. That is a different contest, decided by whether your content can be found, understood and trusted.
By the time a visitor arrives from an AI answer they have already accepted a recommendation. They are warmer than a search visitor — and quicker to leave if the page does not back up what the assistant told them.
This is not a future problem to prepare for. For most markets, the questions that matter already return a confident answer naming specific businesses. The only useful question is which of them are named, and why.
What we examine
Every one of them is a question about your website as it stands today — not a checklist we apply regardless of what we find.
Can the systems reach your content at all? We check indexation, crawlability, rendering, and whether the pages that carry your best answers are actually available to be retrieved.
Does each page state its answer plainly, in the opening lines, in language a model can lift without guessing? Vague, padded or buried answers are the single most common reason a competent business goes uncited.
What does your site say that nobody else could have written? First-hand experience, specific numbers, real examples and named expertise are what give a retrieval system a reason to prefer you over a recycled summary.
Is it obvious, everywhere you appear, that you are one specific business doing specific things in specific places? Contradictory descriptions make models hedge — and a hedged answer usually drops your name.
Semantic HTML, a sane heading hierarchy and accurate structured data. Not a magic switch, but each piece removes a reason for your content to be misread or skipped.
Which businesses are currently surfacing for your questions, what they have that you do not, and which of those gaps are realistically worth closing first.
Answers worth being known for
These are the kinds of prompts an assistant receives today. Open the type closest to your business to see what it takes to be part of the answer.
These are answered from pages that state service area, response times and typical pricing plainly — not from a homepage that says the firm is “trusted” and stops there.
Answered by pages that explain the actual process and show genuine expertise. Generic “we are passionate about law” copy is invisible to these questions.
Answered from product and category pages that carry real specifications and comparisons, rather than a description copied from the manufacturer.
Answered by original reporting, testing and opinion. Summaries of other summaries are exactly what a generative engine has no reason to cite.
Answered from consistent location detail across your site and your business listings. Contradictions between the two are where local visibility quietly dies.
Answered by documentation, comparison and pricing pages that are honest about limits. Buying committees now start these questions in an assistant, not a demo call.
These are illustrative prompts, not a promise about any specific result. What they show is the shape of the work: clear questions, plainly answered, with enough evidence behind them to be worth quoting. See what we point out for every website type
Honest limits
This is the part of the market where the most nonsense is sold. We would rather lose a sale than add to it — so here is exactly what is outside our control.
If a provider guarantees you a fixed position inside ChatGPT, they are describing something they cannot deliver. There is no setting for it, and no tool that has access to those systems.
What we can do is remove the reasons you are currently passed over. That materially improves your odds, and it is the only honest version of this service.
No outside party controls which sources a model retrieves or quotes, and the systems change without notice. Any agency promising you a fixed position in ChatGPT is describing something they cannot deliver.
Google has stated plainly that its Search systems do not use llms.txt, and that there is no special schema.org markup required to appear in generative AI features. We will build those files if you want them for other systems, but we will not sell them to you as a ranking lever.
Google's own guidance is that its AI features are rooted in the same ranking and quality systems as ordinary search. That means the work is unglamorous: clearer answers, better evidence, cleaner structure. There is no shortcut being withheld from you.
Visibility work is partly a waiting game, and some changes take weeks to show. You will get an honest account of what we changed, what we expected, and whether it happened — including when the answer is not yet.
How the work runs
AI answers shift continuously, so a single audit is stale almost immediately. This is what happens every week instead.
We record where you currently stand: which of your key questions surface you, which surface competitors, and what the systems appear to be reading when they answer.
We identify what is actually blocking you — an unstated answer, thin evidence, an ambiguous entity, a structural problem — rather than applying a generic checklist.
You receive specific, prioritised instructions: this page, this passage, this change, in this order. Written so you or your developer can act without interpretation.
We go back to the same questions and check whether anything moved. Some changes register quickly; others take weeks. Either way you are told which.
Each week starts from what we already know about your site and your market, so the recommendations get sharper instead of restarting from scratch.
If your pages cannot be crawled, indexed or understood, none of the rest matters. When we find that, we say so and start there — even though it is the least exciting part of the work.
We review your website by hand before accepting anything, including a first pass at your current AI visibility. Nothing is charged at that stage.
Request Website ReviewQuestions about AI search
Including the uncomfortable ones about what cannot be guaranteed.
When somebody asks ChatGPT, Perplexity, Gemini or Google’s AI Overviews a question, the answer is assembled from pages those systems can retrieve and trust. Being visible in AI search means your website is one of the sources behind that answer. It is a different job from ranking on a results page, because the model is choosing what to say rather than listing where to go.
No, and you should be sceptical of anyone who says they can. No outside party controls which sources a model retrieves or quotes, and the systems change constantly. What we can do is remove the reasons you are currently being passed over: unclear answers, weak evidence, an ambiguous brand identity, or pages the crawlers cannot read. That materially improves your odds. It does not buy you a guarantee.
No, but the two are not opposites either. Traditional SEO is the foundation — a page that cannot be crawled, indexed and understood will never be retrieved, no matter how well written it is. On top of that foundation, AI visibility work is about what happens next: whether your pages state answers plainly, whether they carry evidence worth quoting, and whether a model can confidently attribute them to your business. We do both, and we start with whichever is holding you back.
No, and you should not. Google’s own guidance is that its AI features are rooted in the same ranking and quality systems as ordinary search. Treating them as separate budgets usually means doing half a job twice. Our weekly work prioritises whichever is more limiting for your site that week — and quite often the same change improves both.
We track the questions that matter in your market and check, week by week, which businesses are being surfaced for them and whether that changes. Alongside that we watch the things we can measure properly: whether your pages are indexed, whether assistants can retrieve your content, whether your business detail is consistent everywhere it appears, and whether the visitors arriving from AI answers behave differently from the rest. We report what we see, including when the answer is “no movement yet”.
We will create them if you want them for other systems, but we do not sell them as a ranking lever. Google has stated plainly that its Search systems do not use llms.txt and that there is no special markup required to appear in generative AI features. Anyone charging a premium for a text file is selling you something Google ignores. Our effort goes into content quality, structure and clarity instead, because that is what actually gets retrieved.
Send us your website. We will review it by hand, tell you honestly where you currently stand in AI answers, and only then discuss a subscription.