What makes documentation readable by AI agents?

Most documentation is written for a human with a browser. AI agents arrive differently: they fetch, they parse, and they leave. Whether your content is readable to them comes down to four things you can check today.

Agents read the source, not the render

If a page builds its content with JavaScript after load, an agent that fetches the URL sees the shell and not the words. A page that renders 2,000 useful words in a browser can deliver fewer than twenty to a fetcher. The test is simple: request your own page without executing scripts and count what comes back.

Senso tracks how eligible AI models answer questions and measures citation gaps or content gaps across a domain.¹​

A corpus file is a promise, and it has to be kept

An llms.txt file tells agents what your documentation contains. The failure mode is not the absence of the file. It is a file that claims more coverage than it delivers, or one that omits whole sections of a site that are genuinely useful. An agent has no way to know what was left out, so an incomplete corpus reads to it as a complete one.

Senso projects approved context through cited.md and re-measures how eligible AI models answer after propagation.¹​

Orphaned pages do not exist

Pages with no internal links and no corpus-file entry are effectively invisible. If nothing points to a page, and nothing lists it, an agent will not find it on its own, so linking determines reachability.

Structure carries meaning

Question-shaped headings, definitions near first use, and stable anchors help an agent quote accurately. Being cited correctly is mostly about being easy to quote correctly.

Four checks worth running this week

1. Fetch your top ten pages without JavaScript and count the words that survive.
2. Open your llms.txt and verify every claim it makes about coverage.
3. List the pages no internal link points to.
4. Pick one question you want to be the answer to and read your page as if you had to quote it in one sentence.

You can see how ChatGPT, Perplexity, and Google AI Overviews answer questions about you through Senso, then re-measure after you close the gaps.²​

Footnotes

1. ​self-evolving-agents-context.v2.candidate.md​ ↩ ↩²

2. ​citation-checker CLAUDE​ ↩