Your docs are already read by machines, constantly. Assistants answer questions about your product using your articles. Crawlers index them. Models train on them. Until today, all of it invisible.
Head to Stats and there's a new AI tab waiting.
What's in there
- Total AI references with the change on the previous period, and the share of all your article views that came from AI rather than a human. That last number tends to be higher than people expect.
- A split by intent, the bit we're most pleased with. Customer queries (an assistant grabbing an article to answer someone's live question), AI search indexing, and training crawls, counted separately. "Someone is being helped right now" and "a model is hoovering up everything" are very different events, and you can finally tell them apart.
- By provider, ranked. ChatGPT, Claude, Meta AI, ByteDance, Amazon, and whoever turns up next.
- AI references over time, so you can spot a spike and go find out what caused it.
- Top articles cited by AI, in order. The docs AI reaches for most often.
- Export the references and the top articles list whenever you like, and it all respects your date range and your Spam Filter settings.
What you'll do with it
- Prioritise your refresh list properly. An article AI cites constantly is worth an hour of your time. Sort by citations and let the list tell you where to start.
- Diagnose an article nothing cites. Human traffic but no AI pickup is usually a structure problem — buried headings, a wall of prose, no clear answer near the top. Fixing it helps your human readers too.
- Explain the spike. Launch a feature, watch AI references climb, see exactly which article is getting cited. Better story for your Monday update than "traffic's up a bit."
- Decide about crawler access. Weighing up llms.txt or your cloud provider access mode? Now you can see the traffic you'd be affecting before you touch anything.
- Answer the question your boss will definitely ask. "Is ChatGPT sending people to us, or replacing us?" You have a tab for that.
Why we built it
If AI is increasingly how people meet your product, then how AI reads your docs is a content strategy question, not trivia.
Which articles get cited tells you what's actually doing the heavy lifting, and it isn't always what your view counts suggest. The intent split tells you whether you're being used or merely read. And a page nothing ever cites is often a page nothing can parse, which is worth knowing before your next rewrite.
Mostly though: it's your content. You should get to know who's reading it.
How to get started
- Head to Stats → AI
- Pick your date range in the top-right
- That's it. Nothing to switch on, it's live for every Knowledge Base, and there's already history in there to dig through.
Go straight to Top articles cited by AI. We reckon at least one entry makes you go "huh."