When someone in Surrey asks ChatGPT, Gemini or Google for a roofer in Guildford or a dentist near Woking, the answer is a shortlist of three to five names. If yours is not one of them, the enquiry never arrives and you never know it existed.
Who it is forEstablished local service businesses — installers, builders and roofers, private dental and aesthetics, physio, chiropractic and osteopathy, independent estate agents — with a reputation offline that the assistants have not caught up with.
Ten blue links gave everyone on page one a chance. An AI answer gives three to five names a chance, and nobody else any at all. Nothing shows in your analytics either way, because nobody clicked anything.
The question is asked in full sentences, with a town in it. “Who is a good roofer near Godalming that does slate?”
Three to five businesses, drawn from what the engines can read about you and corroborate elsewhere.
Everyone below it is invisible to that buyer, on that day, for that job. You never learn the enquiry existed.
The engines are not guessing from memory any more. With web search on, every answer is assembled from live pages in a few seconds. Steps three and four are where your website either helps or gets in the way.
Service, town, and usually a qualifier — emergency, family, NHS, Victorian, same week.
The engine runs one or more searches of its own and pulls back a handful of pages.
Only pages it can actually read count: plain HTML, crawlers allowed, content that says what you do and where.
Firms that are clear, specific and corroborated by directories and reviews get named. Vague ones do not.
The pages it leaned on are listed. Being one of them is a stronger position than being mentioned in passing.
An illustrative sample. The firm, the competitors and every number below are made up, and a real report says the date and the domain it was run for. What is real is the shape: the count by engine, who was named instead, and the one gate that explains most of the gap.
Named in 2 of 30 answers overall. Cited as a source in 0.
| Business | Times named |
|---|---|
| Surrey RoofcareAll five engines | 24 |
| Guildford Roofing CoChatGPT, Gemini, Perplexity | 19 |
| A1 Roofing GodalmingGemini, AI Overviews | 11 |
The report counts both, separately, because they are different positions and need different work to reach.
The engine has heard of you — usually through a directory, a review platform or a mention on somebody else’s page — and puts your name on the shortlist.
The engine read your site while writing the answer and lists it as a reference. The buyer can click straight through, and the answer is built on what you actually say.
A gate is a fact about your site that keeps the engines from reading, trusting or naming you. The check tests all seven and shows the first three it found, ordered by effect. The audit works through the rest.
Enter the domain and watch it run. It reads what your site already publishes, works out who you compete with, and puts the questions a buyer would actually ask to five answer engines with live web search on.
Homepage and up to five pages: services, towns served, trading details. Plus robots.txt, sitemap and schema.
The businesses the engines already recommend across your service area, so the report has names in it.
Local, comparison, problem-led and trust questions, in the words your buyers use, with your towns in them.
ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews, each with web search on. Every answer kept.
Named, cited, who instead — and the specific gate that explains why it is not you.
The free check tells you where you stand. What follows is priced on the page like everything else here, rolling monthly with one month’s notice.
25–40 real buyer searches across five engines, captured as evidence and scored against three named local competitors. All seven gates, with the change that opens each. A one-page 90-day plan you could run with anyone.
The 90-day plan worked month by month, in order of effect: town pages, service copy, schema, crawler access, and the directories and review platforms the engines cite. Each change recorded against the question it should move.
The same questions re-run against the baseline the audit set, across all five engines, so you watch the count of times named move and see who is being named in your place. Reported in plain words, no dashboard to log into.
Nobody controls what an AI answer says, so exactly as with rankings, we guarantee the evidence, the dates and the plan, never the mention.
Yes. No card, no call first. The headline result is shown as soon as the run finishes. The full report, with every question and every answer, is sent to the email address you give us, and that is the only thing we ask for.
Your homepage and up to five of your public pages, plus robots.txt and your sitemap, exactly as a visitor or a crawler would see them. No login, no analytics access. Your site content is not kept beyond the run. The raw answers from each engine are stored so the report can show you the evidence.
No, and nobody honestly can. Nobody controls what an AI answer says, including the companies that run the engines. What the check gives you is evidence and a dated baseline, and what the audit gives you is the specific changes that make you eligible to be named.
Enough to show a pattern rather than a fluke, and few enough to run in about four minutes. Ten are local service questions, eight are comparisons, six are problem-led, and six are about trust. They are generated from what your site says you do and where you do it, in the words a buyer would use.
One run per domain every seven days. If someone has already checked your domain this week you will see that result. Three runs per email address a month. The audit and the retainer re-run the same questions against your baseline, so you can watch the count move.
About four minutes. You will see how many of the 30 answers named you, who was named instead, and the first three gates. Then decide whether the audit is worth £850 to you.
Would rather talk it through first? Book thirty minutes, weekday afternoons. Or go straight to the audit.