Which engines does each plan ask?
The trial asks the standard panel: ChatGPT, Perplexity, Claude and Google AI Overviews. A paid plan starts with the same 4 and picks from the price card's 4, ChatGPT, Perplexity, Google AI Overview and Claude, each priced per question; at least one is always on. Each engine is asked the way it reaches buyers, with web search where the product itself searches.
| Engine | How the question is asked | What answers |
|---|---|---|
| ChatGPT | As a logged-out visitor on chatgpt.com, read through a data provider. If that read fails, the OpenAI API answers instead, and the stored answer records which one it was. | Whatever model chatgpt.com serves a logged-out visitor that day |
| Perplexity | Perplexity's own API, which searches the web before it answers. | Sonar |
| Claude | Anthropic's API with the web search tool switched on. | Claude Haiku 4.5 |
| Google AI Overviews | The question is searched on Google and the AI Overview on the results page is read, through a search data provider with a browser read as the fallback. | Google's AI Overview, as shown on the page |
How do I see which model gave an answer?
- Open AI Visibility and go to the Prompts tab.
- Open a prompt, then open one engine's answer.
- Response Details shows the full answer, its citations and the model version that produced it. When an engine does not report one, it says Model not recorded.
Providers sometimes change a model without notice. Because the version is stored on every answer, a shift in the numbers can be checked against a model change instead of guessed at. How models and samples feed each figure is on Methodology.
Can I add more AI engines?
Yes. Google AI Mode, Gemini, Microsoft Copilot, Grok, DeepSeek and Meta AI are enabled on a call (Book a demo), each priced per question like the other engines, or on Managed, where our team does the work. Extra engines never change the standard-panel headline, so a number stays comparable whether or not they are on.
Which country and language are the questions asked in?
The one you choose. On Check what we found, Asking AI as a buyer in sets the country the questions are asked from, and Answer language sets the language. The report then says which country the answers on screen were asked from, so you always know whose answers you are reading.
What happens when an engine gives no answer?
When Google shows no AI Overview for a question, that is stored as an empty answer. Empty answers are kept in the record and in exports but left out of mention rate and share of voice, so a missing Overview never reads as a loss. A read that fails is recorded as a failure and retried, never as an empty answer.
