How answers are collected and counted

What one answer is

One answer is one prompt asked to one engine, once. A scan of 5 prompts across ChatGPT and Perplexity records 10 answers, and 10 answers are deducted from the monthly allowance, whether or not you were cited.

Which engines

ChatGPT and Perplexity, on every plan. Each engine receives your prompt text verbatim, with no instructions about your product.

Not included. Google AI Overviews has no API; Google’s terms forbid analysing Gemini’s grounded results; Claude is built and tested and will be switched on later.

What is recorded

For each answer: the engine, the prompt, a short snippet of the answer, and every cited URL with its position (1 is the first source cited). A URL counts as yours when its host matches your domain, as a competitor’s when it matches a competitor you added, and as third-party otherwise.

How the report counts

Source ranking sorts URLs by number of citations, then by average position. Your share is the number of citations pointing at your domain divided by all citations in the scan. Coverage is, per prompt and engine, the best position at which your domain was cited.

Change between two scans lists sources that appeared or disappeared and prompts where you gained or lost a citation. Two scans are two points in time; the report does not call that a trend.

Failed and partial scans

If an engine cannot be reached, the scan stops. Answers recorded before the stop are kept and count towards the month, because the engines were really asked. A scan with no recorded answers is marked failed and costs nothing.

What is not measured

Traffic, positions in web search, and the wording of answers beyond the stored snippet. AnswerGround does not promise a position or an improvement; it records what the engines did on the day of the scan.