- Data window
- The four totals cover every stored answer of customer teams since 2026-02-09, the first answer PromptScout kept; they are single numbers. The shares, sites and trend use only the previous calendar month's answers (UTC), a chosen subset, and the trend is that month's last four seven-day weeks. The worked example shows the runs of the previous calendar month only. There is no way to page through or download the stored answers.
- Engines
- ChatGPT, Gemini, Google AI Overviews, Perplexity and Bing Copilot. An engine or a week with answers from fewer than three customer teams shows a gap, never a zero.
- What is counted
- A stored AI answer is one engine's reply to one monitored question on one run. It names a brand when PromptScout stored at least one brand or competitor name for it. A citation entry is one source link on an answer; the same site in two answers counts twice. Google's encrypted redirect links name no site and are left out of the sites.
- Who is shown
- Only aggregates over customer teams: PromptScout's test accounts and internal subscriptions are left out. Every figure from last month needs answers from at least three customer teams; with fewer, the page shows the month's answer count only. A site appears only when answers of at least three customer teams cite it, and never when it is a customer's own domain. The worked example comes from PromptScout's own workspace. No customer team, brand or question appears on this page or in its data response.
- Where the numbers come from
- One read-only database query computes the counts, shares, sites and trend at most once a day, and each figure shows when. The same data is published at /api/public/visibility-lab.
- What is not checked
- Whether an answer is correct, complete or current. The benchmark records what each engine named and cited, not whether it should have.