Tools
The first two gates, checked live.
The first two gates can be read from outside a site, so here they are, free: two checks on crawl access and two on what the index would keep. Each states what it observed, what it inferred, and what it does not claim. Gates 3 to 5 need a prompt panel and cannot be read by a tool on a web page.
Crawl check
Gate 1. What the site returns to each of nine named search and AI crawlers, and what its robots.txt says to each. Finds the disagreement between the file and the edge.
Index check
Gate 2. What the index would keep from a URL, read from the HTML a crawler receives before anything renders: canonical, directives, text, structured data, sitemap.
Check the index signals →Sitemap check
Gate 2, from the sitemap's side. What the sitemap declares, and whether ten declared URLs fetched as Googlebot are as declared, or redirect, error, or withdraw themselves.
Check the sitemap →Identity check
Gate 1, from the logs. Paste a log line, or a sample of up to 200: was each request that presented a crawler's name inside that operator's published address ranges, and does reverse DNS forward-confirm it?
Check a log line →Why only the first two gates
A tool on a web page can make a request and read a response. That is enough to see gate 1, and enough to read the signals of gate 2 in the HTML. It is not enough to see whether an engine kept a page, which is Search Console's reading, and it is nowhere near enough to see retrieval, citation or influence, which take a frozen panel of at least 120 prompts, five repetitions per engine, and a noise floor measured first. A single prompt asked once carries roughly 44 points of error. Any tool that shows a number for gates 3 to 5 from one request is showing noise with a decimal point.
All five gates, in the form a client receives, are applied to this site in the self-audit, with every reading dated and gates 3 to 5 listed as not measured.