What the audit checks, and what it deliberately does not
Every check comes from a Google Search Central guide. The things Google says it ignores are never flagged.
The Site Audit is built from Google's own documentation at developers.google.com/search, read page by page. Each factor in the table names the rule it applies. Nothing on the list is an industry convention dressed up as a Google requirement, and several things other tools flag are left out on purpose because Google has said they do not matter.
Checked from the crawl
- robots.txt, parsed the way Google parses it: per-crawler groups, Allow and Disallow, longest rule wins, the 500 KiB cap. Blocking Googlebot-Image or Google-InspectionTool is flagged; blocking Google-Extended is a choice and is not.
- noindex behind a robots.txt Disallow. Google can only read a noindex on a page it may fetch, so this combination never works.
- An element that does not belong in <head> (an img, iframe or div). Google stops reading the head there, so everything after it is lost.
- Sitemaps within Google's limits: 50,000 URLs and 50 MB, absolute URLs, no nested index, lastmod that makes sense. changefreq and priority are reported as ignored, because Google ignores them.
- Redirect chains and temporary redirects on internal links, soft 404s, meta refresh, and paginated pages canonicalised to page 1.
- Favicon (square, at least 8x8 and ideally over 48x48, in a format Google reads; SVG is not), Organization logo at least 112x112, Discover images at least 1200 pixels wide with max-image-preview:large, article dates that agree with the visible date, and VideoObject thumbnails.
- Structured data for 30 rich-result types against Google's feature guides: required properties, exact limits and enums, markup that describes something the page does not show, and the difference between a product snippet and a merchant listing.
- Cloaking (a different page for Googlebot), sneaky redirects (a different destination for Googlebot or for visitors from Google), and mobile parity (the phone page missing content, structured data or directives the desktop page has). A sample of ten pages is fetched again for these.
- Core Web Vitals from Google's CrUX field data through PageSpeed Insights, on the homepage and the pages with the most Search Console clicks: LCP 2.5 s, INP 200 ms, CLS 0.1. The Lighthouse lab score is shown but never judged, because Google ranks on field data.
Reported but never scored
Some rows say Not scored. They are things worth a look that Google has said are not ranking signals: heading order, link titles, long alt text, high link counts, duplicate IDs, security headers, hidden text by inline style, sitemap tags Google ignores, retired robots directives, and the recommended (not required) structured-data properties. They never move the health score.
Not checked, because Google says it does not matter
- Meta keywords. Google does not use them.
- Word count. Google says there is no magical word count, minimum or maximum.
- Keywords in the domain or URL, and the choice of top-level domain.
- rel=next and rel=prev. Google stopped reading them.
- A title length limit. Google states none; long titles are cut to fit, not penalised.
- llms.txt for Google. Google Search does not read it. Other AI crawlers do, which is why the platform still writes one.
Google's Indexing API accepts JobPosting and BroadcastEvent pages only. Submit to Google sends other URLs through IndexNow and Bing and marks them skipped for Google; for a single page, use Index status to request inspection.
The source every check cites.
Related guides
Crawl any domain live and grade its technical SEO in about two minutes.
Every fix is a before/after diff you control, nothing goes live until you say so.
Pull real clicks, impressions, positions and the queries you rank for.
We’ll get back within one business day.