Why generic prompts fail
A generic "improve my website" prompt usually returns a clean-looking list with weak prioritization and no measurable targets.
AuditHQ already gives you the missing context: current score, exact issue titles, and severity. Using that context changes output quality immediately.
The minimum inputs for a useful prompt
Include project name, target URL, current headline score, and top high-priority issues from the latest run.
Ask for phased execution windows (24 hours, 7 days, 30 days) plus effort sizing and expected impact for each action.
Require measurable acceptance criteria for every task (for example: Lighthouse threshold, contrast ratio pass, or reduced unused JS bytes).
Where to run it in AuditHQ
Use homepage quick audit for early triage, then create a project to unlock detailed history and the copy-ready AI fix prompt.
Project-level prompt generation stays tied to your latest report, so when issues change, the next prompt changes too.
Quality control before execution
Reject advice that is not tied to your listed issues, lacks expected impact, or has no verification step.
After shipping fixes, run another audit and compare the new score and issue list before starting the next prompt cycle.