Every page we score runs through 36 checks in 6 categories, with weights summing to 100. The engine is deterministic — no AI, no network calls, no randomness. The same page always produces the same score, which is what makes two scores comparable at all.
Scores use the same letter bands as the audit: 97+ is an A+, 90+ an A, 80+ a B, 70+ a C, 60+ a D, and below 60 an F.
The six categories
| Category | Checks | Weight | What it measures |
|---|---|---|---|
| Keyword & Intent | 8 | 25 | Whether the page targets a real search and satisfies the intent behind it: the keyword in the right places, at a sane density, matched to the page type — and, for planned pages, whether it keeps the plan's secondary keywords and AI prompts. |
| Depth & Readability | 8 | 20 | Whether there is enough substance, structured so a human can read it: length against page type, paragraph and sentence shape, scannability, and whether the figures it quotes are grounded. |
| Structure | 6 | 15 | Headings, hierarchy, and whether the page is organised rather than merely long. |
| Meta & SERP | 4 | 15 | Title and meta description as they will render in results, and whether they are unique against your other pages. |
| AEO / GEO | 5 | 13 | Whether an AI search engine can lift a clean answer out of the page: direct answers, structured data, citable passages. |
| Links & Media | 5 | 12 | Internal links out and in, external references, images and alt text. |
Keyword & Intent carries the most weight, so it moves the score most: a well-structured page that targets no real search still scores poorly.
Read the results
Each check returns a pass, a warning, or a fail, with the specific fix. Open a page's score to see all 36 checks, not only the failures — the passes tell you what not to break.
Two numbers are stamped on every stored score:
- Engine version: which revision of these checks produced the score.
- Playbook version: which revision of the SEO guidance was active.
If a score changes and you didn't touch the page, check whether the engine version moved.
Where the score comes from
Three places produce scored pages, and they all use this one engine:
- The audit scores pages it crawls.
- Article generation → in Skyscraper mode scores each draft and revises until it clears the bar.
- The content refresher → scores a live page before and after.
A generated article and an audited page are scored by identical rules. Nothing is graded on a curve for being AI-written.
Limits and usage
- Scoring is free and instant. It runs locally with no model call.
- It measures the page as fetched. A page that renders its content in JavaScript will score as thin, because that's what a crawler sees too.
- It can't judge whether your claims are true or your advice is good. It measures form, not correctness.
If something isn't working
A page scores much lower than it reads. Check word count and internal links first — those are the two most common causes.
A page has no target keyword. Keyword & Intent checks degrade rather than fail outright, but the score will be capped. Map the page in the keyword map.
The score dropped after a re-run with no edit. Compare engine versions, then check whether the page's siblings changed. Duplicate-title detection compares against your other pages, so a new page can pull an old one down.