What one analysis measures
- Lab Core Web Vitals: Largest Contentful Paint with the element and its phases, Cumulative Layout Shift, and Total Blocking Time as a lab proxy, scored with a Lighthouse-style curve.
- A network table captured through the browser dev-tools protocol: transfer versus decoded bytes, protocol, TLS, cache, third parties by registrable domain, compression, mixed content and console errors.
- An image-delivery audit against Lighthouse thresholds, an axe-core accessibility check, an Observatory-style security-header grade, and a table-driven tech-stack fingerprint that records the evidence for each match.
Read the numbers honestly
The lab run is an unthrottled desktop load from a data centre, not a throttled mobile run like a public field tool, so the numbers are legitimately better and are not directly comparable. Interaction to Next Paint is a field metric and is not measured here unless field data is configured. Every section carries its own provenance and a data-quality banner, so a measurement that could not be taken shows as unavailable with a reason, and a bot-walled or thin page served to a data-centre address is flagged rather than scored as if it were the real page.
Accessibility here is a check, not a certificate
The accessibility section runs axe-core and reports an impact-weighted score, and it never sets a conformance claim. It is an indicative check that finds a subset of issues, not a WCAG or EN 301 549 conformance statement. For a real audit with a VPAT, use the accessibility module, which is built for exactly that.
Export and re-read
Every analysis is saved and can be re-opened, and a report exports to Markdown or JSON. Because the result is stored per analysis, you can compare a page against its earlier self after a change.
Limits
- It analyses one page, not a whole site. There is no crawl, so do not read a single result as a site-wide audit.
- Some enrichment is conditional on configuration: field-metric interaction data, keyword or search-ranking data, and domain-authority ranks depend on operator-provided keys or files and degrade honestly when they are absent.
- The composite health score is a deterministic, rule-based number. It is not a trained model, and the tool says so; do not read it as AI scoring.
Questions people ask about this
- Does it crawl my whole website?
- No. It analyses exactly the one URL you give it, from one page load. A whole-site crawl is a different job and this is not it.
- Is the accessibility result a conformance claim?
- No. It is an axe-core check that finds a subset of issues and always sets the conformance claim to false. For procurement, use the accessibility audit and VPAT flow instead.
Source: /documentation/website-analysis, updated 2026-09-18. Answers are checked against the shipped product on that date.
Found something out of date? The people who build the module answer the contact form on the home page, and corrections to this guide are dated above. Tell us.