Engines covered
KlientBrain AI Visibility currently tracks your brand across these AI engines:
| Engine | What is asked | Location |
|---|---|---|
| ChatGPT (app) | The ChatGPT consumer app | country |
| Perplexity | Perplexity's API (Agent API, fast preset) | country |
| Google AI Overviews | Google search results, including the AI Overview when one is shown | country |
| Google AI Mode | Google's AI Mode answer | country |
| Gemini | The Gemini app | country |
| Claude | Anthropic's API with web search (Claude Haiku 5.5) | approximate city |
| ChatGPT (API) | OpenAI's API with web search (GPT-6 Luna). An Enterprise option | approximate city |
Each plan starts with a default set of engines. See Plans, checks and limits. You can switch engines on and off for a project in Settings, under Tracking.
Pinned models
The model that answers each run is recorded with the answer and shown in the report. Where an engine lets us choose the model, it is pinned, so the same model answers every time. For engines that read a consumer app or Google's results, the app or Google chooses.
Vendors sometimes change the model behind an engine without notice. When the model that answered differs from the previous full run, the app marks it: the trend chart shows a dashed line, the engine shows "New model" instead of a change, and the report notes it. How we handle model changes
How location is set
Location changes what an engine answers, so it is set for every run and written in the report:
- Approximate city, where the engine supports it (Claude and the ChatGPT API).
- Country, for Perplexity, Google AI Overviews, Google AI Mode, the ChatGPT app and the Gemini app.
- Prompt wording only, for Gemini's API model, which has no location setting.
Settings, under How we measure, lists the location method for each engine on your project.
When an engine gives no answer
Where each answer comes from, and whether it can come back empty:
- ChatGPT (app): the answer the ChatGPT app shows. The app sometimes gives no answer.
- Perplexity: Perplexity's answer from its API, with its own web search.
- Google AI Overviews: the AI Overview at the top of Google's results for the prompt's search. Google shows one for some searches only, so on some runs there is none.
- Google AI Mode: Google's AI Mode answer for the search. It sometimes gives no answer.
- Gemini: the answer the Gemini app shows. The app sometimes gives no answer.
- Claude: Claude's answer from Anthropic's API, with web search.
- ChatGPT (API): ChatGPT's answer from OpenAI's API, with web search.
An engine with no answer for a prompt on a run is not counted as a miss. The answer is stored with its reason and left out of your visibility and the other figures on the reports, so they are worked out from the answers that came back. It still counts as a check used. The same prompt may get an answer on the next run.
What you see:
- Open the answer: it reads No AI answer shown., then the reason, for example "No AI Overview for this search".
- On Overview, the "Based on" line under the source types counts these with failed calls, as "failed and excluded".
- If an engine stops answering, or gives no answer far more often than before, the likely causes of a change list it as No AI answer shown, with a count such as 12 of 12.
Failed calls are stored, shown as This answer failed., and left out of the report figures too. Unlike an answer with no AI answer shown, a failed call does not count as a check used.