Know whether AI names you, and how often

• Your organic traffic is falling and you do not know how much of that demand is being resolved inside a chat, with nobody clicking.
• You asked ChatGPT and it named you. You asked again and it did not. Two queries decide nothing.
• Your manager or your client wants a number, and all you have is a handful of screenshots.

One mention is not data. Three hundred counted answers are.

1. 1

   You define the questions

   Six intent families: brand, discovery, comparison, decision, product and trust. Write them, import them, or start from the ones the system proposes for your sector.

2. 2

   They run against the engines

   Through each model's official API, with web search on and the country you choose. Every answer records its fidelity level, and none is presented as if it came from the consumer app.

3. 3

   What came back gets counted

   Each mention is located in the text, attributed to a brand and classified. Homonyms are separated by a dedicated classifier, and exclusions are counted apart so you know how many answers dropped out.

• Presence rate and share of voice, always together. A hundred per cent share can be three mentions in three hundred answers, and you read that on the same line.
• The full answers stored, with the mention highlighted and the link the engine cited. It is the evidence you show when someone says they get something else.
• The comparison with the previous sweep, engine by engine, to see whether what you changed did anything.

• SEO leads who already feel the drop and need a number where today there is only a suspicion.
• Agencies reselling the data who need to explain the method before they are asked.
• Brands with homonyms that need to separate their mentions from the other company with the same name.

The 14-day trial brings two engines and 25 questions, no card. Starter widens the set and the cadence; Pro and Agency add Claude, Perplexity and Google AI summaries, plus repetitions and several brands. Spend has a hard cap on every plan: you set it and the system does not cross it.

See the full pricing

Is this really ChatGPT or is it the API?

It is the official API of the model behind ChatGPT, with web search enabled. It is not identical to asking in the app: the app adds its own context and history. We say so here, we say so inside the product, and every answer carries its fidelity level. Anyone who tells you it is the same thing is lying.

How many questions does it take for the number to mean anything?

It depends on what you want to claim. To know whether you exist in a category, 20 or 30 questions per engine already show it. To claim you improved on last quarter you need more repetitions, because the variation between two runs of the same model is not zero. The dashboard always shows the denominator so you can judge it yourself instead of trusting us.

My brand shares its name with another company. Does that not break the count?

It would if nothing were done. There are two layers: an exclusion list you define, and a classifier that decides whether an ambiguous mention is you or the other company. Discarded answers do not vanish: they are counted apart and the dashboard tells you how many, because a denominator that silently shrinks is worse than no denominator.

Where are the answers stored and who sees them?

Data is hosted in Spain, on our own server. The questions travel to AI providers in the United States so they can answer them, and we say so rather than hide it. Inside your account each person sees what their role allows, and client accounts only see their own brand.