The AI industry has the same visibility problem it sells the cure for

There is a particular irony in AI companies being invisible to AI. Ask an assistant for "a vector database that runs on-premise" or "an OCR API that handles handwritten forms" and the shortlist is short, repetitive, and usually a few years out of date — because the answer comes from documentation and forum threads that were text, not from the launch sites that were animation.

Documentation is your best crawler surface, and you already have it

If your docs are plain HTML or markdown, a crawler reads them properly. This is why some small tools punch far above their marketing weight in AI answers: their reference pages state precisely what the thing does, in what language, with what limits.

The gap is that documentation answers how, not whether. Nothing on a docs page says "use this when you need X and not Y" — and that is the sentence a model needs to recommend you to somebody describing a problem.

Write the positioning a model can quote

• What class of thing is it. Vector database, OCR API, feature store — the ordinary category word.
• What is the boundary. Runs on-premise, no GPU required, handles up to N documents a second.
• What it is not for. Naming the wrong use case makes the right one unmistakable.
• Pricing shape. Open source with a paid tier, usage-based, seat-based. Say which.

What is listed here

Each company below has a listing page and a markdown twin, linked from this article, the category feed and llms.txt, with an open count of how many times each named AI crawler fetched it — from server logs, unedited.