When AI assistants answer a question, they cite sources. Over the past year, those citations have quietly become a new ranking signal that nobody quite knows how to game. We pulled citation data from a sample of 8,400 queries across ChatGPT search, Claude, and Perplexity to see who’s actually getting cited — and the pattern doesn’t look like classic SEO.
The biggest finding: encyclopedic and government sources dominate. Wikipedia, .gov domains, and major university .edu pages account for a disproportionate share of citations relative to their search traffic. A press release on a .gov site can outrank a far better-written Forbes article for the same query, because the chatbot defaults to authority signals classic search engines downweighted years ago.
Second pattern: forums and Reddit threads punch far above their weight, especially for practical “how do I” questions. Six of the top 50 cited sources in our sample were Reddit subreddits. For consumer product queries specifically, Reddit was cited more often than any single review site.
Third pattern, and the strangest one: smaller niche sites with strong topical authority get cited at rates traditional SEO would never predict. A self-published industry blog with 40 articles can show up alongside Bloomberg in citations, because the chatbot is reading topical depth, not Moz-style domain authority.
The implication for publishers is uncomfortable: the playbook that worked for Google — long, comprehensive, keyword-tuned articles — isn’t necessarily what wins citations from AI. What wins instead looks like authoritative writing in narrow domains, the kind most SEO advice has been telling people to abandon for a decade.