A list of failed checks is not a to-do list. Prescriptions turn each finding into a concrete instruction for a specific page — and verification confirms you actually fixed it.

From finding to prescription

For each failed or warned check in the website diagnosis, the dashboard generates a prescription that tells you:

  • which URL to change (typically your homepage for homepage-level checks, or robots.txt / llms.txt for crawler files)
  • what exactly to add or change — e.g. the JSON-LD block to embed, the llms.txt content to upload, the robots.txt lines to remove
  • why it matters, tied to the severity (P0/P1/P2) so you can order the work

Prescriptions also feed the recommendations board, where each item moves through Suggested → To-Do → Done — useful when several people share the work.

The verification loop

Fixes to your own site don't need to wait for the next weekly scan:

  1. Deploy your fix (add the schema, upload llms.txt, unblock the crawler…).
  2. In the dashboard, mark the prescription done.
  3. Trigger the re-check — the audit re-fetches the affected URL immediately and re-runs the specific check.
  4. The check flips to pass if the fix is live, or stays failed with fresh detail if something is still wrong (a typo in the JSON-LD, a cached old robots.txt, etc.).

This loop is deliberately fast so you can iterate: fix, re-check, fix again, all within minutes.

What verification does not do

A passing check means your site is technically legible to AI systems — it does not guarantee mentions. Model behavior shifts slowly; content and citations (Citation Gap) drive the rest. Judge mention-rate impact over weeks of monitoring, not the next scan.