AI Brand Check — is your brand in the AI's answer? Free.

Ask ChatGPT, Perplexity, Claude or Gemini one buyer question, three times, and see whether your brand is in the answer, how often, in what position, who else is named, and which pages the engine cited. Free. No API keys, no signup beyond your Apify account.

by Desvela — the free step before AI Brand Watch.

What you get

One record per check:

Field	What it says
mentionRate	Fraction of the three answers that name your brand (0, 0.33, 0.67, 1)
avgPosition	Where your brand appears among brand + competitors (1 = first)
competitorRates	Same rate for each competitor you listed, or for the ones inferred from your brand
ownDomainCitedRate	Fraction of answers that cite your own site (needs domain)
topSources	The pages the engine cited, most repeated first
answers	One excerpt per answer, around your brand when it appears
verdict	One sentence: "ChatGPT mentioned Notion in 3 of 3 answers to 'best all-in-one workspace for teams'. Competitors named: Monday.com in 3/3, Asana in 2/3."
nextStep	The input JSON, already filled in, to run the full measurement with AI Brand Watch

Leave question empty and the check infers one buyer question and up to three competitors from the brand name (one small model call, on us). Give the domain when the name alone is ambiguous: on a bare "Otterly" the inferred question was about otter plushies, not AI visibility software; with otterly.ai it is not.

How it counts, and why three

AI engines are non-deterministic: ask the same question twice and the brand set can change. One answer is a snapshot chosen by luck. Three answers give you a rate, which is still a small sample, but it already tells 0/3 from 2/3, and that is the difference between "the AI does not know you" and "you are in the running".

Failures are not zeros. If the engine does not answer one of the three attempts, the rate is computed over the answers that came back and failedSamples says how many were lost. If none came back, measured is false and there is no rate at all. A cell we could not measure is not a cell at zero.

We call the engines' APIs with web search on. API answers correlate with, but are not identical to, each consumer product. We always measure the same way, so two checks are comparable with each other.

What this is not

It is one question, one engine, three samples, once. It is not a measurement of your AI visibility: visibility is a rate over the 5-10 questions your buyers actually ask, across the engines they use, tracked over time so you know when it moves. That is what AI Brand Watch does, and every check ends with the input to run it: the question you just tested, the competitors, promptSetAction: keep so the question set stays frozen, and autoGenerateQueries: true so the actor adds the rest of the category. A full first run costs about $1.66 (pay per check, no subscription); add an Apify Schedule to make it weekly.

Limits

Every check is paid from our own API keys, so there are two caps: 3 checks per account per day, and a monthly budget across all users. When a cap is reached the run does not fail; it returns a quota_exhausted record that says which cap, when it resets, and how to run the uncapped actor instead.

Input

{
  "brandName": "Notion",
  "domain": "notion.so",
  "question": "best all-in-one workspace for teams",
  "engine": "openai",
  "competitors": ["Confluence", "Monday.com", "Asana"]
}

Only brandName is required. engine is one of openai (ChatGPT, default), perplexity, anthropic (Claude), gemini.

Real example

Run on 2026-09-18, brand Notion, question inferred:

ChatGPT (OpenAI API, web search on) mentioned Notion in 3 of 3 answers to "best all-in-one workspace for teams". Competitors named: Monday.com in 3/3, Asana in 2/3.

avgPosition 1.3, ownDomainCitedRate 0 (the engine cited comparison sites, not notion.so), topSources led by two "X vs Y vs Z" articles. That last detail is the kind of thing a rate over time would tell you about: who the engine trusts to describe your category.