An audit that tells you what it could not measure
Thirty-four checks across your metadata, headings, content, links, images, indexing, security and schema, each finding carrying one of five severities. The health score lists the dimensions behind it and marks the ones the provider did not return, so a number you act on is a number you can take apart.
Proof in the workflow
The score shows its own arithmetic
On-page SEO carries 0.40 of it, performance 0.20, accessibility and best practices 0.15 each, security 0.10. When the performance provider returns nothing, that dimension is marked unavailable and the remaining weights are renormalized, so the headline does not silently lose twenty points to a category nobody scored.
It knows when it is reading an empty shell
A page whose text is drawn by JavaScript arrives at the crawler as a document with no headings and almost no words. That gets its own finding, and the checks that depend on body copy are withheld for that page, so you are told the page was unreadable.
Findings open in the surface that fixes them
A missing-schema finding opens the schema generator, an orphan page opens link suggestions, a thin page opens the editor.
What this landing is built to rank for
Thirty-four checks across eleven categories
Metadata, headings, content, links, images, indexing, crawlability, security, structured data, mobile and accessibility. Titles and descriptions that are missing, duplicated or the wrong length; missing H1s and multiple H1s; broken internal links, redirect chains and redirect loops; missing alt text and missing image dimensions; noindex pages, canonical mismatches, missing HTTPS and missing HSTS.
Two ways to run it
Point it at a single URL for a page audit with Lighthouse performance, accessibility, best-practice and security scores attached. Or crawl the site: up to 500 URLs discovered through the sitemap and internal links, up to 100 of them fetched and analyzed.
Orphan pages, or a straight answer about why not
The crawl records every internal link it follows and lists the pages nothing points at. It declines to run that pass on a truncated crawl, on a site whose link graph came back with no edges, and on one where any page arrived as an empty shell, because each of those produces a list of orphans that are not orphaned.
Re-crawl closes what you fixed, and only that
Run it again and findings that no longer appear are marked fixed with a timestamp. The pass only covers pages this run actually crawled, and it excludes the checks the run did not perform, so a shell page or a skipped orphan pass leaves those findings open at their original age.
How it works
Audit a page, or crawl the site
One URL gets the full page audit with Lighthouse attached. A site crawl walks the sitemap and the internal links from the homepage out.
Read the score with its rationale
Five dimensions, each with its weight, its raw score and a note when the provider had nothing to give.
Work the findings by severity
Critical, high, medium, low and info, each finding naming the page, the value it found and what it expected instead.
Re-crawl and read what closed
Findings that no longer appear on a crawled page are marked fixed and timestamped. Past runs stay in history and reopen with their full result.
Best fit use cases
SEO teams in rapid release cycles
Re-crawl after a release and the closed findings list what it fixed.
Agencies handling technical cleanups
Each client site keeps its own crawl history, its own open findings and its own fixed count.
Teams shipping a JavaScript front end
The shell check tells you which pages serve no text to anything that reads the response as sent.
Supporting content
FAQ
How many pages does one crawl cover?
It discovers up to 500 URLs through the sitemap and internal links, then fetches and analyzes up to 100 of them per run. A crawl that hit the cap is recorded as truncated, and the orphan pass is skipped on those runs, since a truncated crawl has not seen every page that might link to the rest.
Does it check robots.txt and sitemap coverage?
The crawler reads the sitemap to discover URLs, but it does not currently file findings about robots.txt rules or about pages missing from the sitemap. The 34 checks it does run cover metadata, headings, content, links, images, indexing, crawlability, security, structured data, mobile and accessibility.
What happens when Lighthouse cannot score a page?
The dimension is marked unavailable with the reason attached, its weight goes to zero, and the remaining weights are renormalized so the overall score stays a real average of what was measured. Without the performance provider, the score is the on-page result on its own and says so.
How does this differ from a static site crawl?
Every finding carries a severity and the page it was found on, each run is stored and reopenable, and a later crawl closes the findings that no longer appear. Findings also open in the surface that fixes them.