Audits and reports
AI summary
What it does, what we send, and why it never changes the audit.
AI summary
The AI summary sits above the tabs on the report screen. It reads your audit and writes a short paragraph in your language telling you which fixes to do first.
What it does
The summary is a plain-language executive brief: which findings are most important, why they matter, and roughly which one to start with. It does not add new findings — it re-tells the ones the audit already produced.
What we send
Only two things are sent to the AI model:
- The text of your audit findings — the same rule titles, plain-word explanations, and fix strings you can see on the screen.
- Your affected URLs, so it can reference them.
We do not send Search Console data, screenshots, your billing info, or anything about other sites on your account.
Why it never affects the audit itself
The audit is deterministic — same input, same output. The AI summary is a separate step that runs after the audit finishes, using the finished snapshot as read-only input. Turning the summary off has no effect on any rule’s bucket.
Regenerating
Click Regenerate on the summary card to run it again. Each regeneration uses one summary credit (see the AI summaries cap in Plans, limits & credits). If you hit the cap the card shows a paywall notice; unpaid retries do not spend credit.
Choosing a model
The summary uses a small, low-cost Claude model by default. Some findings — very long lists — are shortened before sending. If shortening happens, the card shows a "shortened" note.