Which AI assistants does PromptMentions track?
Nine AI engines: six are asked directly; Google’s AI answers and the ChatGPT app are captured as they answered. Copilot on Bing is paused.
The engines
PromptMentions tracks nine AI engines, and all of them are on every plan:
- Asked directly: ChatGPT (shown in the product as OpenAI, the company that makes it), Perplexity, Claude, Gemini, Grok and Meta AI.
- Captured from the search results page: Google AI Overviews and Google AI Mode.
- Captured from the ChatGPT app: ChatGPT app (captured), shown next to OpenAI and never mixed with it.
- Paused: Copilot on Bing. Our data supplier is returning the wrong results pages, so it cannot be ticked for now. Answers collected before the pause stay visible.
The bar at the top of Visibility > Answers shows each engine as connected or not connected for your workspace.
When you start a check you tick the engines to ask, up to four at a time, and a schedule also runs up to four. Under Engines for this client on a schedule you choose which ones.
An engine that is not connected collects nothing. It is shown as not connected and no zero is filled in for it, so a gap in the results always means “not asked”, never “not mentioned”.
How answers are collected from ChatGPT, Perplexity, Claude, Gemini, Grok and Meta AI
Each question is sent to the engine through the official connection its maker provides for software, with web search switched on, for the country and language you chose. The answer is saved whole, with the engine, the exact model, the wording of the question, the place and the time, and the sources the answer listed.
Claude is asked without extended thinking and for an answer of about 220 words at most, so its answers here can be shorter than the ones the Claude app gives.
Is that the same as what I see in the ChatGPT app?
Close, but not identical. The app adds things of its own: your past conversations, your settings, sometimes a different model. So an answer collected here shows what the engine says to a neutral asker, not what one particular person saw on their phone. That is why every figure is labelled with where it came from, and why an answer from OpenAI is never presented as a capture of the app.
To see the app itself, tick ChatGPT app (captured) (below). Its answers stay in their own row, next to OpenAI, and the two are never added together.
ChatGPT app (captured)
OpenAI offers no official connection for what the ChatGPT app shows, so a specialist provider asks the ChatGPT app your question, for one place at one time, and PromptMentions saves the whole answer with the sources ChatGPT cited and the model it reported. Every such answer is labelled as captured from the ChatGPT app.
- It runs on the same question sets and schedules as the other engines, and each answer uses one answer from your plan.
- It is available for question sets in 43 countries: every country you can choose in PromptMentions except China and Hong Kong, which the capture service does not offer.
- It is not your own signed-in chat: no past conversations or personal settings are involved. ChatGPT decides for itself whether to search the web.
- A capture can take up to two minutes, so a run that includes it takes longer to finish. A capture that fails or comes back empty is shown as failed and left out of your rates.
Google AI Overviews, Google AI Mode and Copilot on Bing (paused)
Google and Microsoft offer no official connection for these answers, so they are collected differently. A specialist provider loads the search results page for your question, for one place at one time, and PromptMentions saves the AI answer shown on that page with the pages it listed as sources. Every such answer is labelled as captured from a results page.
- They run on the same question sets and schedules as the other engines, and each check uses one answer from your plan.
- They are available for question sets in ten countries: the United Kingdom, the United States, Ireland, Canada, Australia, Germany, France, the Netherlands, Spain and Italy.
- Many searches show no AI answer at all. Such a check is reported as “no AI answer on the page” and is left out of your rates. It is never counted as “not mentioned”.
- Copilot on Bing is the answer Bing shows above its results, not the Copilot chat app.
- Copilot on Bing is paused: our data supplier is returning the wrong results pages. We will switch it back on when that is fixed. Until then it cannot be ticked or scheduled, schedules that include it run without it, and nothing is charged for it. Earlier Copilot answers stay visible; checks whose page was for a different search are shown as “Wrong results page”.
Explorer > Variants & engines also has More AI engines, where you can check one question once with Check once.
Related articles
- How often are results refreshed, and how do I run a check?Results are refreshed when you start a check or on the schedule you save, from every 7 days to every hour depending on your plan.
- What do mention rate, citation rate and share of voice mean?Brand mentioned, site cited, share of voice, position and win rate, explained in plain words.
- How do I see which sources ChatGPT and Perplexity cite?See which websites ChatGPT and Perplexity cite when buyers ask about your category, or a client’s, and find the ones that mention competitors but not you.
Still need a hand?
Use the chat button in the corner of the page, send us a message or write to support@promptmentions.com. A person reads every message.
Looking for something else? See what the product does, pricing, how each figure is measured and how your data is protected.
Last checked against the product on 17 November 2026. All help articles