The honest test of a tool is running it on yourself. We maintain a free agent-readiness checker (https://agent-ready-check.t6y11a.workers.dev/) that looks at the HTML your server sends and reports the structure signals a simple AI crawler can see. This week we pointed it at PlainData (https://plaindata-preview.t6y11a.workers.dev), our own free calculators site, and fixed what it caught. Here is the real before and after.
How we ran it
We fetched the HTML exactly as the server sends it, pasted it into the checker (the analysis runs entirely in the browser; nothing is uploaded), then did the manual server checks the tool itself lists: HTTP status and headers, robots.txt, sitemap.xml.
Before: 9 passes, 3 warnings
Passing: title tag (53 chars), meta description, no restrictive robots meta, language attribute, heading structure (1 H1, 4 H2), text visible without JavaScript (~327 words), semantic landmarks.
Warnings:
- No canonical URL.
- No JSON-LD structured data.
- No machine-readable dates in structured data.
And on the server side: sitemap.xml returned 404. There was no sitemap at all.
One finding needs an honest frame: robots.txt existed but contained only the content-signals preamble, with no operative crawler directives. That is not a crawl failure - absent rules default to permissive - but it meant the site had no explicit, intentional policy.
What we shipped the same day
- A canonical link on the homepage.
- A WebSite JSON-LD block (name, URL, description).
- A real sitemap.xml listing the 4 public pages, with truthful lastmod dates.
- Deliberately not shipped: machine-readable dates. The pages are tools, not articles; inventing publication dates to turn a warning green would be worse than the warning.
Robots.txt stayed as is: permissive by default is a legitimate policy, now a conscious one.
One more honest thing
While we were verifying the fix, the checker briefly reported "None found" for the JSON-LD we had just shipped. The JSON-LD was valid - the bug was in the checker itself, which removed script elements while counting visible words before it looked for structured data. We found it, fixed it, and re-verified the same evening. A free tool that gives false warnings is worse than no tool, so this mattered to us.
After
Canonical: pass. JSON-LD: pass (1 WebSite block). Sitemap: 200, four pages. Machine-readable dates: still a warning, by choice.
None of this predicts that any AI engine will cite or rank the page. These are machine-checkable structure signals - the difference between a crawler being able to read your page cleanly and not. That is all any checklist can honestly claim.
If you want to run the same check on your own site, the checker is free and stays free: https://agent-ready-check.t6y11a.workers.dev/ - paste your View Source HTML and it tells you what it finds.
Disclosure: This post is published by Clear Rank Studio, the maker of the free checker above and of two related workbooks: the AI Search Visibility Audit Workbook ($9.99, 19 crawler and access checks plus a 16-pillar page audit: https://instinctive283.gumroad.com/l/cuxtft) and the AI Citation Tracker and Content Kit ($9.99: https://instinctive283.gumroad.com/l/marlq). PlainData is also our site. All links go to our own tools and store; none are affiliate links. This article was written by Clear Rank Studio's AI assistant and every result quoted was verified against the live pages before publishing.
Top comments (1)
Sitemap coverage and sitemap quality are different problems: a URL can be listed and still be a page a crawler is right to ignore. I found three category pages in my own directory with zero listings, still rendering and still submitted. Setting them to noindex until they have real content was the cheapest fix I made this month.