The report that starts the AI visibility conversation with your client

You already do the work. Local SEO, reviews, content, digital PR — the things most likely to bear on whether an AI engine names your client are largely things you already do. What you probably don't have is the diagnostic: a specific, checkable, client-branded artifact that shows a business owner that ChatGPT is recommending someone else by name.

That artifact is what AskedAbout sells. Run it on a client, put your logo on it, and let the engines make your argument for you.

What we found in 153 free checks of real local businesses

This is our own first-party data — every number below comes from checks we ran against live AI engines, not from a survey or a vendor report. Read it with its label: these are free checks run in June and July 2026, as counted when this page was written on 2026-07-12 — The score is the blended free-check score of that period. The original runs do not carry the method metadata now recorded for paid audits; their question panel should not be inferred from today’s free-check form. The paid full audit uses a larger question panel, four engines and a separate discovery score, so a full-audit score should not be compared with these averages.

Read the bottom row slowly. In that set, the average local business was absent from more than half of the sampled answers — and that count includes the question that handed the engines its name. In 65 of 153 cases, a single named competitor was named more often than the business itself. Across the set, the engines named 552 distinct rival businesses — specific recommendations, every one of them going to somebody else.

Most clients have not seen that page about themselves. It gives the conversation something specific to start from: what the engines said, and who they named.

Why the work is mostly off the client's own site

In a 21-business cut of our home-services audits, we looked at which sources the engines actually cited when recommending a local company. The answer reframes the work: the client's own website was cited in only about a fifth of answers. The engines were reading third-party evidence — directories, review sites, forum threads — far more than the homepage.

That cut suggests AI visibility is unlikely to be settled by rewriting a client's homepage once: much of what the engines read sits off-site. What it does not show is which change will move an answer, how long that takes, or that any particular engagement is needed — a citation count says what an engine read, not why it named a business. The audit gives you the observations; the scope of work is your judgement and your client's.

How agencies use the 5-pack

1. 1Run a free check first. Use the form on this page with a real client's name and city. Sixty seconds, six AI answers, no card. You'll see immediately whether there's a story to tell.
2. 2Buy the $249 Agency 5-pack — five full audits, $50 each. No subscription, no seat fee, no annual contract.
3. 3Redeem a credit per client. Each full audit requests up to 25 buyer questions × 4 engines × 3 samples — up to 300 AI answers — and the report states how many were requested, completed and failed per engine. It reports mention rate per engine, share-of-voice against every competitor the engines name, the sources they cite about that market, and a prioritized fix plan of suggested actions.
4. 4Put it in the deck. The report is yours to re-brand and hand over. We are not in the room, we don't contact your client, and nothing in the deliverable requires us to be.

A report that names the competitors the engines recommended, with the answers quoted, gives a client something checkable to react to. Get the 5-pack →

How the audit works

We query the official APIs of ChatGPT, Perplexity, Gemini and Claude with two kinds of question. Most are discovery questions a buyer might ask without knowing the client — “best med spa in Scottsdale”, “who should I call for emergency plumbing in Austin” — and they never contain the client's name. The current full-audit panel has 22 discovery questions and 3 brand questions that do use the name (“is it legit”, “reviews”); the report marks each one. Older reports retain their original panel and identify missing method metadata. The two kinds are scored separately — discovery visibility and brand recognition — because a brand-name question tells you nothing about discovery. Because AI answers are nondeterministic, we ask each question three times per engine and report a frequency, not a screenshot, along with how many answers were requested, completed and failed. These are API answers: what a person sees in a consumer chat app can differ.

An agreed one-client pilot

For an agreed pilot, three credits from the $249 five-pack cover one client: a baseline, an unchanged repeat, and a repeat after the agency records its changes. Two credits remain for ordinary audits. You approve the same 25 questions and the service, buyer and market before running them. Each wave requests 300 answers across four provider APIs; the report identifies completed, missing and failed answers. The comparison includes a CSV with counts and method details. Before-and-after differences do not establish that your changes caused them. We confirm the written scope and delivery readiness before accepting a pilot; read the proposed scope.