Projects » CITED
Finds which sources an AI answer engine actually reads for your category, then opens the pull request that fixes your gaps.
Built at the c0mpiled 11 Startup School Hackathon on 24 July 2026 at Transpose Platform in San Francisco. Won the Hexclave track — we built the agent and sold it to one of the hackathon’s own sponsors within three hours.
The premise
Your site ranks on Google. That is increasingly not where your buyers are asking.
Roughly 80% of the URLs cited by AI answer engines don’t rank in Google’s top 100 for the same query, and only about 12% appear in the top 10. Two consequences follow that most teams haven’t internalised:
- Your Google rank tells you almost nothing about whether ChatGPT recommends you.
- The pages that do get cited are mostly pages you don’t own — listicles, review sites, comparison pages, forum threads.
Why this one and not the twelve monitoring tools
The answer-engine monitoring category is crowded and well funded. Every one of those tools stops at a dashboard: here is your visibility score, good luck. CITED closes the loop — it produces the actual content change and opens the pull request against your repo.
The pipeline takes a site URL and a repo URL, queries the answer engine, aggregates citations by domain, scores visibility, diffs your content against the gap keywords, generates metadata rewrites and a gap-targeting page, and opens a PR.
A detail I liked
It ships with a fixture mode that renders the entire dashboard from a local JSON file with no keys and no network. Venue wifi at hackathons always dies. Ours didn’t matter.
- Runs N category queries against an answer engine and captures every cited URL
- Scores visibility as cited queries over total queries
- Diffs your content against gap keywords
- Generates metadata rewrites plus a gap-targeting page, then opens the PR