Guides/AI answers and insights
Insights: metrics and conversations
Insights shows you what happens after you publish: who reads your docs, what visitors ask, and where your docs come up short. It is the feedback loop that tells you what to write next.
Open Insights in the dashboard sidebar. Everything reports on the last 30 days.
Metrics
Four tiles up top:
- Article reads: views on your docs site, split between humans and AI agents. Agent traffic (coding tools and crawlers reading your
.mdandllms.txtendpoints) counts separately, so you see both audiences. - Questions asked: how many questions visitors asked the AI.
- Answered rate: the share of questions that got a cited answer.
- Couldn't answer: questions where your docs had nothing to offer.
Below the tiles, two daily charts track reads (agents vs humans) and questions (answered vs couldn't answer) over time.
Two panels close the loop:
- Most cited articles: the articles the AI leans on most when answering. These are your load-bearing docs; keep them sharp.
- Couldn't answer: the latest questions that stumped the AI, each linking to its conversation.
Conversations
The Conversations tab lists every AI conversation with its full transcript: what was asked, what was answered, which articles were cited, and which answers came up empty. Filter to "Couldn't answer" to see only the gaps.
Conversations are anonymous and kept for 30 days. See Data retention and privacy.
Close the loop from your coding tool
Insights are available to your agent too, which is where they become actionable. Ask:
What are my docs missing? Check recent unanswered questions and draft articles to cover them.
Your agent can pull the same report (reads, answered rate, most cited articles, unanswered questions), read full transcripts, and go straight to drafting the missing article. That is the whole loop: publish, watch what visitors ask, fill the gaps, publish again.