Configure a site audit.

Crawl breadth — stop after this many distinct pages. The server enforces your plan's cap whatever this field says, and duplicate URLs of the same page are detected and skipped without spending your budget. A capped run is flagged partial in the report, so it never reads as a whole-site score.

Pages audited simultaneously — each is a browser tab. 3 is the polite default; 8 is fastest but heavier on memory and harder on the target site.

Also audit what pages do: modals, tab panels, accordions, and the errors a form shows when required fields are empty. Each costs a full evaluation plus a reload — 3–4 roughly triples scan time. 0 = off.

Catches failures that only appear once the layout narrows — content hidden or overlapping at a breakpoint, contrast against a responsive background. Roughly doubles scan time.

Real screen-reader output as supplementary evidence on the reading-order and status-message checkpoints. Needs the capture pipeline installed on this machine (NVDA, or an attached Android device) — otherwise it does nothing. Off by default.