Free · AI visibility audit
Is AI telling your buyers about you?
We ask Claude, ChatGPT and Gemini four questions a real B2B buyer would ask — and show you exactly what they say about your company. No fluff. No generic checklist. Just the transcript.
A decorative panel showing one example transcript: a buyer question, the answer an engine gave, and what that answer shows. The text is illustrative, not a live call. Your own report shows every real answer verbatim.
An example of what an unaided-discovery answer looks like — not a live call, and not your data. Your own run is a dated API sample, with every answer shown verbatim.
How it works
Four prompts, three engines, one honest read.
We ask the APIs
A small set of buyer-style questions goes to Claude, ChatGPT and Gemini through their APIs. Some questions name you; the discovery questions never do.
We classify each answer
Was your brand named? Recommended? Was the description accurate? Answers that did not arrive stay out of the count.
We show the evidence
Every answer, verbatim, with the model, the tool mode and the date. A sample of a few calls — not what every buyer sees.
You get a fix list
Each recommendation points at the evidence behind it, with effort and confidence. No promised score jump.
What we measure
What the audit actually measures.
Four axes, each a share of the answers we actually received. The free run measures three — grounded search is opt-in, and a report never shows it as a zero.
Brand recall
The question names you. Does the engine describe you, and does it get the description right?
Unaided discovery
The question never names you. Do you come up at all when a buyer asks who to use in your category?
Grounded search
Web search or grounding on, counting citations that return your own domain as a source. Opt-in — not part of the free run.
Technical readiness
Deterministic checks on your own pages for the signals those engines depend on. No AI call involved.
Why this is not an SEO report.
Ranking rewards pages that compete for a query. Retrieval rewards pages that state a verifiable fact an engine can lift and attribute. A page can rank on page one and still be invisible to an assistant answering the same question.
So we do not send you a keyword list. We send you the sentences three assistants actually produced about your company, the names they returned instead, and the specific pages worth rewriting first.
Method
How we measure.
- A dated API sample: each report records the day the calls were made. AI answers change between runs.
- Model and tool mode are named on every call (for example web search on or off). We do not replay the consumer apps.
- Small sample: a handful of calls per engine. It shows a pattern, not a ranking.
- A provider error or timeout is reported as unavailable, never as a zero. A measured zero is shown as 0.
- Grounded search citations are a separate, opt-in axis. When it is not part of a run, the report says “Not included in this run” — not zero citations.
- The score covers the axes a run actually includes, and it always says which. The free run measures three, so its number is not comparable to one that includes grounded search. If an axis is attempted and fails, the score is withheld instead — a failure is not the same as an omission.
Already seen your results? Contact us to go further.
Built by operators, not integrators. Montreal, Miami, Paris.
FAQ
Questions, answered.
What is a GEO audit?
A GEO audit measures whether AI assistants name and describe your company. We send a small set of buyer-style questions to Claude, ChatGPT and Gemini through their APIs, classify each answer, and check your site for the technical signals those engines rely on. The free run measures whether the engines name you, plus technical readiness.
How is the score calculated?
Each axis is a share of the answers actually received: brand recall (the question names you), unaided discovery (it never does), grounded search (web search or grounding on, counting own-domain citations) and technical readiness (deterministic page checks). Grounded search is opt-in and is not part of the free run, so it is reported as “Not included in this run” rather than zero. The overall number covers the axes a run actually includes, and the report always says which: the free run measures three, so its number is not comparable to one that includes grounded search. An axis a run never offered narrows the score; an axis that was attempted and failed withholds it instead. A provider error or timeout is reported as unavailable, never as a zero.
Which AI engines are tested?
Claude (Anthropic), ChatGPT (OpenAI) and Gemini (Google), through their APIs. Each report names the exact model used and whether web search or grounding was on. It is a dated API sample of a few calls per engine, not a replay of the consumer apps.
Does the free audit check who AI cites?
No. A citation is provider citation metadata returning your own domain as a source, and that requires grounded search, which is opt-in and not part of the free run. The free audit measures whether the engines name your brand — a mention — and checks your technical readiness. Naming your brand is reported as “Mentioned”; “Cited” is reserved for a real source citation.
Is my data stored?
We store the business name and domain you confirm, your contact name and email, and the report, so we can deliver the report and let our team follow up. Newsletter consent is a separate option that is off by default. We do not sell or share your data with third parties.
Why does a report say “Unavailable”, “Not measurable” or “Not included in this run”?
They mean different things. “Unavailable” means we tried and did not receive enough answers to report a rate. “Not included in this run” means that axis was never attempted — grounded search citations are opt-in and are not part of the free run, so it is not a measured zero. “Not measurable” is the overall number when an axis that was attempted did not come back: a failure withholds the score rather than narrowing it. A true measured zero is shown as 0.
Do I need to be technical?
No. The report is written in plain English with specific recommendations, each tied to the evidence that produced it, with effort and confidence. No SEO jargon. No technical prerequisites.