How answers are collected
Cleotic asks each AI engine your prompts the way a person would ask in the engine's own app. This page explains what happens on every run, so you know what the evidence in Answers, AI citations, and Content Studio represents.
What the engine is told
Each prompt is sent exactly as you wrote it. The engine is also told three things its app would already know about the person asking. These are today's date, the market of the prompt's segment, and the segment's language.
The engine isn't told it's being monitored, and it's never asked to mention a brand. Engines that can think before they answer use a light amount of thinking on every run, which keeps runs comparable over time.
Each engine uses its own web search
Every engine searches the web with the same search its own app uses.
- ChatGPT uses OpenAI's search.
- Claude uses Anthropic's search.
- Gemini uses Google Search.
- Muse uses Meta's search.
- Perplexity uses its built-in search.
Cleotic only sends a run where the engine's own search is available. If it isn't, the run fails rather than quietly using a different search engine, so an answer never reflects search results the engine's app wouldn't use.
Google AI Overviews and Google AI Mode come straight from Google's results page, so you see what Google showed for the question.
The engine decides whether to search and how often, just as it would in its app. Some questions get answered without any search at all.
The whole answer is kept
Answers are stored in full, however long they are. Some engines write short progress notes before they answer, such as "I'll research the best options for you." Cleotic keeps only the final answer, which is what you'd read in the app.
What each engine reports
Engines differ in what they tell us about how they found an answer.
| Engine | The searches it ran | Pages it read but didn't cite | How it cites sources |
|---|---|---|---|
| ChatGPT | Yes | No | Sources linked to the sentence they support |
| Claude | Yes | Yes | Sources listed with the answer |
| Gemini | How many, but not what for | No | Sources linked to the sentence they support |
| Muse | Yes | No | Sources linked to the sentence they support |
| Perplexity | No | No | Numbered markers such as [1] and [2] |
| Google AI Overviews and AI Mode | No | No | Google's source cards and inline links |
You'll find the searches and pages on each answer. See Responses. For what counts as a source, see Citations.
Why an answer can differ from the app
Answers vary from run to run, even for the same question. An engine's app may also know things about you that a run doesn't, such as your account, past chats, and exact location. Look at trends across many runs before reading too much into a single answer.