Claude's citations: 64% brand sites, zero Reddit
Every engine has a personality, and Claude's is now measured: it is the one major AI engine where your own website, not Reddit, is the main character. Otterly's June study classified 379,321 Claude citations across 16,406 domains, and the profile it found is so different from ChatGPT's that treating "AI search" as one channel stops making sense the moment you see the numbers.
Where Claude's citations actually go
Inside that sliver of social, LinkedIn takes 43.1% and Reddit takes nothing at all. A separate cross-engine sample measured Claude's forum citations at roughly 0.1%, and an academic analysis of Claude's health citations found 97.8% pointing at institutional sources such as government bodies, hospitals and professional associations. The pattern is consistent from three independent angles: Claude trusts organizations over crowds.
A different animal from ChatGPT
The overlap numbers are the strategic core of the study: just 13% of Claude's cited domains are also cited by ChatGPT, and at the URL level the overlap falls to 4.2%. Claude also refuses to form a club: its top ten domains account for only 9.5% of citations, and 32.9% of cited domains appear exactly once. Compare that with the cross-engine picture, where a short platform list dominates everything, and you get two channels wearing one label. Optimizing Reddit threads for ChatGPT does nothing here. Publishing reference-grade pages on your own domain does everything.
One more property matters for planning: the mix barely moves. Across the weeks of the study, no source category shifted by more than about a percentage point. What wins Claude citations this month should still win them next month.
What it means for your content
Claude skews toward research, comparison and B2B decision queries, exactly the traffic that converts. And it is the one engine where the work you already control, your product pages, docs, pricing pages and first-party research, is the primary citation pool rather than the long shot. The playbook is the entity-first brief with the third-party layer turned down and the first-party layer turned up: reference-grade pages, evidence in every passage, entities stated cleanly. This week, ask Claude the five buying questions in your category and read what it cites. If your own site is absent from the one engine that prefers brand sites, that is the cheapest gap in AI search to close.
