Reading the AI report
The AI Visibility page answers one question: is AI reading you, and is it sending anyone. It is built to be read top to bottom.
The hero numbers
Three counts for the period:
- AI citations — total live AI-answer fetches. This grows as AI systems reference you more.
- Pages cited — how many distinct pages got at least one AI-answer fetch.
- Confirmed click-throughs — real humans an AI answer actually referred to your site. This is the payoff metric, and it carries the accent for that reason.
These count live AI-answer fetches only. Index sweeps and training crawls are excluded here — see The bot taxonomy for why.
The trend chart
AI activity over time, split by category: AI answers, indexing, and training. The AI-answers series is the one to watch; the other two show raw crawl volume for context.
The generated report
A short written read of the period, derived from your own data — no model call:
- A lead finding: usually your most-fetched page.
- Actions — findings that imply something to do, such as a page AI fetches a lot but that no human ever clicks through from, or a page an assistant fetched that returned a 404.
- Notes — informational findings that do not need action.
There is a "Copy for AI" button that renders the same report as markdown, so you can paste it into an assistant.
The per-page table
One row per page, sorted by fetch count. Each row shows fetches, the platform breakdown, a trend versus the previous period, a confidence level, and an intent class — informational, research, commercial, or navigational — inferred from the path. Commercial and navigational intent are highlighted, because those are the pages tied to an action.
Expand a row to see the inferred prompt families and recent sessions.