Usability scan

A usability scan scores a page against a benchmark of usability checkpoints and files what it finds in the research graph — so a usability problem becomes evidence you can cite and link, rather than an opinion in a review.

Run a scan

Three ways in, all producing the same kind of findings:

• Research → Create → Usability scan. Paste a public URL, pick a viewport (desktop, tablet, or mobile), and optionally narrow the scan to a few categories.
• The browser extension. Open the Usability tab in the side panel and scan the tab you are on. Findings are pinned onto the page itself, coloured by severity.
• An agent. Ask Crux to run a usability scan on a URL, or call run_usability_scan from a coding agent over MCP.

What gets checked

The benchmark is a corpus of checkpoints across twelve categories:

navigation and findability · search and filtering · forms and data entry · system status and feedback · error prevention and recovery · content and information architecture · trust, credibility and privacy · accessibility and inclusive design · performance and perceived speed · mobile and touch ergonomics · visual hierarchy and legibility · task completion and friction

Every checkpoint cites at least one openly-licensed or publicly-citable source — WCAG and the WAI-ARIA practices, the GOV.UK Design System, the Web Almanac, published research. Paywalled guideline text is never reproduced.

Server scan or live scan

The two entry points see different amounts of the page, and the scan says which it used.

A checkpoint that cannot be judged at the current fidelity is reported as not evaluated rather than passed, so a thin scan never reads as a clean bill of health.

What a finding says

Each finding names the checkpoint it failed and what good would have looked like, then adds:

• a severity — P0 blocks the task, P1 causes significant friction, P2 is measurable drag, P3 is refinement;
• the evidence on the page, including the element it applies to;
• a suggested fix;
• the sources behind the rule.

Findings that need a human or real users to settle are flagged for manual validation instead of being asserted.

Findings in the research graph

A scan and its findings are research items. They appear in the Research list with everything else, can be filtered and searched, and can be cited by Ask like any other evidence. Link a finding to the experience entity it affects and the problem travels with that page or component — into its Graph tab, into the why-story on your live site, and into whatever your coding agent reads before it starts work.

Each finding carries a status — open, accepted, dismissed, or fixed — so a scan can be worked through rather than just read.

Next

• Experience scan & code sync
• MCP, overlay, and the browser extension
• The research graph