How we score documentation
A strict scale for how easy your docs are to find, read and act on.
Seven areas, one consistent scale
Our scale is strict. Every audit uses the same checks. Search estimates and AI recommendations do not change the score.
We review up to 25 public pages from your sitemap and navigation. Home and collection pages are evaluated as navigation, with summaries; article checks look at the answer. The report shows the earned total when at least one readable page is available. If no pages can be read, we do not create a report.
Crawlability & indexability
- Documentation response failures. Full credit requires HTTP 200 for each attempted documentation URL.
- Public indexing directives. Full credit requires no explicit noindex directive on sampled public pages.
- Search crawler permissions. Full credit requires known Googlebot permission for all sampled paths.
- Canonical declarations. Checks for an explicit canonical declaration; alternate canonical destinations are reviewed separately.
Information architecture
- Readable documentation URLs. Flags opaque identifiers and paths deeper than seven segments.
- Collection navigation. A recognizable breadcrumb or at least two distinct documentation destinations earns full credit; one destination earns half.
- Focused content regions. Checks for a recognizable main, article or documentation body region.
- Descriptive sections. Pages with at least 180 words should expose a descriptive subsection. Short answers need not add sections.
On-page search signals
- Descriptive page titles. A non-generic HTML title with at least three words earns full credit; a shorter topic title earns half.
- Distinct page titles. Checks for non-empty titles distinct within this sample, ignoring case and whitespace.
- Useful search summaries. A meta description with at least eight words earns full credit; a shorter summary earns half.
- Clear main topics. Full credit requires one non-generic main heading; multiple distinct H1s or a generic heading with a descriptive HTML title earn half.
Answer & AI readability
- Readable answer text. Answer text earns 0 below 40 words, 40% at 40–79, 75% at 80–119 and full credit at 120+. Linked indexes are evaluated as navigation.
- Topic-specific answer openings. An opening paragraph of at least 12 words earns half; a recognizable main-topic term earns the rest. This is a lexical proxy, not a correctness test.
- Structured answer support. Ordered steps, a table, code block or descriptive illustration earns full credit; another list or inline code earns half.
Internal linking
- Useful contextual documentation links. Counts distinct documentation destinations with descriptive labels in the body: one for shorter answers, two for answers of at least 350 words.
Reader next steps
- Task-appropriate next steps. Reference, troubleshooting, concept and collection pages earn full credit for a contextual next step or action. Recognized setup/workflow/integration tasks earn half for a contextual docs link and full credit for a recognizable product action.
Content quality signals
- Explanation depth. Depth scales to 180 words for recognized tasks or 120 for other answers, supported by two substantive paragraphs, an example or steps. Linked indexes are evaluated as collections.
- Practical explanation support. Recognized tasks earn full credit for an example/table/descriptive illustration and half for ordered steps alone. Other answers can earn full credit with three substantive paragraphs.
- Distinct extracted answers. Checks for exactly repeated normalized extracted bodies across sampled URLs.
Reading the results
The audit reviews public HTML and respects robots.txt. It does not confirm live Google indexing, AI citations, rankings, traffic, signups or conversions. A readable sample cannot establish every missing page in a library. Word counts, link patterns and answer structure do not test factual correctness.
Search volumes and positions are third-party estimates for the stated market and date. They are separate from Search Console and Analytics measurements. Missing search data is omitted rather than treated as zero.
Recommendations should be checked against your product. HTML size, fetch duration and older dates alone do not deduct points. Missing inbound links apply only to the pages reviewed.
Comparing scores
When enough comparable audits exist, a report saves a rank at generation time. Comparisons use unexpired audits with the same scoring version and exclude the audited domain. Overall and category comparisons each require at least 20 other sites unless the operator has configured a higher or lower minimum of at least five. Historical reports keep their saved comparison, including when none was available.
