Phare

There is no page two.

When someone asks an AI which software to buy, they don't get ten links to work through. A typical answer names 3 products. That's the shortlist, and it's the whole conversation.

Do you know whether you're on it?

You can check your Google ranking in ten seconds. This one you can't check at all.

A search ranking is public. Anyone can look it up, including you.

An AI answer isn't. It's generated once, for one buyer, and then it's gone. Nobody keeps a copy. No dashboard your agency runs is watching it. So most companies don't have a bad score here — they have no score at all.

You don't have to believe AI search is the future to want to know what it currently says about you.

We ask the questions your buyers ask. Then we check the answers.

Real buying questions, in French, across your category — "quel ATS pour une PME de 50 salariés ?", not keywords. We put them to the assistants and read every answer: who gets named, in what order, and what's said about them.

Then we do the part nobody else does. We fetch your own website and check what those answers claim about you against what your site actually says.

Both sides, word for word, so you can check us.

Most of what an assistant tells buyers about you didn't come from you.

Of 18 statements an assistant made about one company we measured — specific enough to check against their own website — only 8 were supported by anything on it. The other 10 appeared nowhere on it.

None contradicted the site outright — which is not reassurance. It means there was nothing to correct, and most of what buyers hear still isn't coming from them.

One company in our sample, anonymised.

The rest came from somewhere else. Review sites, listicles, a comparison page from 2023, a competitor's landing page. All of it shaping what a buyer hears before they ever reach you, and none of it under your control.

We show you every one, with the quote and the source, and where to go to fix it.

Proof, stated plainly

160 AI answers analysed across 4 French software categories · 40 buyer questions each

Read from the measurement database on 2026-11-22. Not one of these numbers is written into a page.

Who this is for

French companies of roughly 10 to 150 people, selling something buyers compare before they buy. If you already care where you rank on Google, this is the same question about a channel nobody is showing you.

We measure how ChatGPT answers your buyers' questions. Claude, Perplexity and Gemini are in progress.

No score is worse than a bad one.

A bad ranking you can work on. A blind spot you can't.

One report, your real category, your real competitors. No login, no trial, no card.

No analytics. No cookies. No third-party requests.