What this does
Crawls a site up to a page cap you set, then produces a prioritised issue list. The crawler does the mechanical half: status codes, titles, meta descriptions, heading structure, canonical tags, internal link counts, image alt coverage, duplicate content signatures. The model does the half a checker cannot — reading each page and judging whether the content answers the query it is targeting.
That second half is what separates this from a lint pass. A page can be technically flawless and still be the wrong page.
What you get
- The crawler config and the scoring prompt
- A severity rubric — what counts as critical, high, medium, cosmetic — with the reasoning behind each threshold
- Google Sheets output you can sort, filter and assign
- A short “first five fixes” section, because a list of 200 issues is not a plan
What it replaces
Not your judgement. It replaces the subscription you open twice a quarter and the afternoon you lose deciding whether 412 warnings is a lot or normal.
Honest limits
It audits what is publicly reachable. Anything behind a login, disallowed in robots, or rendered only after a client-side fetch will be missed or misread — check the crawl coverage number before you trust the list.
It cannot see your rankings, your competitors’ backlinks, or your Search Console data. It reads your site, not the market. Without real query data you are optimising in the dark.
Severity is a heuristic, not a fact. The rubric ships in the file so you can disagree with it and change it.