Depra captures ChatGPT and Gemini answers from their consumer web apps at chatgpt.com and gemini.google.com, so it records the answers real people see there, with ChatGPT's web search switched on. Google AI Overviews are read from a real Google results page, and Perplexity answers come from its Sonar model with live web search, all asked from inside the country you choose.
What happens to each question, step by step?
- 1. You choose the questions. Each tracked question is a prompt, one question your buyers ask AI. You also pick the countries to ask from.
- 2. Depra asks each engine. On schedule, the question goes to each engine from inside that country.
- 3. The full answer is stored. Depra keeps the whole answer text and every website the answer cites.
- 4. Mentions are read. An AI model reads the answer and records which tracked brands it names, in what order, in what tone, with the exact sentence as proof.
- 5. Scores are calculated. Visibility, share of voice, position and sentiment come from those stored answers.
Where are the questions asked from?
Each question is asked from inside the country you pick, at country level. Depra supports 40 markets, listed on which countries and languages Depra supports.
Depra does not target cities, simulate personas or use a logged-in account with chat history. Every check is a fresh question with no personal history, so results can be compared day to day.
What can this method not see?
- ChatGPT with search off: answers are captured with web search on, so answers given with search off are not measured.
- The Perplexity app: Perplexity answers come from its Sonar model, which can differ from what the Perplexity app shows.
- Pages read but not cited: Depra stores the pages an answer cites. It cannot see pages an engine read and did not cite.