Everything-PR's 2026 AI Citation Source Index, produced in coordination with 5W Public Relations, synthesizes six independent studies — more than 680 million citations across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews — into one ranked list of the fifty domains that decide what AI answers say, with Reddit, Wikipedia, and YouTube on top and a companion podcast study showing the transcript is what gets cited.
2026-10-02 · by Moe Ameen
A 2026 research program from Everything-PR — the research and publishing platform run by 5W Public Relations chairman Ronn Torossian, produced in coordination with 5W's AI Communications practice — pulled together six independent citation studies — work from Profound, Semrush, Ahrefs, Search Engine Land, Muck Rack, and Visual Capitalist — into a single AI Citation Source Index. The combined dataset covers more than 680 million individual citations produced by five answer engines (ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews) roughly between August 2024 and April 2026, and ranks the fifty domains those engines reach for most.
The top of the list is a short, stable set of names. Reddit is the single most-cited domain across engines, running near 40% of multi-engine aggregate citation frequency and close to half of Perplexity's top-10 share. Wikipedia is next and is the dominant source inside ChatGPT specifically, where various counts put its share of ChatGPT's top-10 citations anywhere from roughly a quarter to nearly half. YouTube, LinkedIn, and Forbes round out the top five, followed by Amazon, Business Insider, TechRadar, Reuters, and The New York Times. The index reports that the top fifteen domains alone capture about 68% of all consolidated AI citation share — a striking concentration that means a small number of third-party, mostly user-generated or reference surfaces are deciding what AI answers say about entire categories.
A companion study in the same 2026 research series looked specifically at podcasts, which the authors describe as the largest ungoverned citation surface in AI search — millions of shows in the catalogs, but only a few dozen that the engines cite with any regularity. The podcast findings land on a detail creators tend to miss: an engine cannot hear audio, so the citable artifact is the written transcript. Shows that fall out of the top ranks tend to fail on transcript access and extractability, not on the quality of the conversation.
The through-line across both studies is that AI-citation authority lives overwhelmingly off a brand's own website. The surfaces the engines trust most are communal (Reddit), encyclopedic (Wikipedia), and multimedia-with-a-transcript (YouTube and podcasts) — none of them your marketing site. The specific percentages move by study, engine, and query category, so treat them as direction and priority rather than fixed targets; the ranking order has been the stable part.
The index is a map of where to show up, and it quietly splits into two columns. Some of the top-cited surfaces are earned — you cannot buy your way onto Wikipedia or astroturf your way up Reddit, and you should not try. But two of the names in the top five, YouTube and LinkedIn, are platforms you publish to directly, and that is the opening a creator can act on this week. [Kompozy](/) publishes natively to both: point it at a long recording, a webinar, or a podcast appearance and it generates [Clipped Shorts](/glossary/clipped-short) and [Persona Shorts](/glossary/persona-shorts) for YouTube and the kind of expert-voiced posts LinkedIn's AI citations overwhelmingly favor — so you are producing content that lands squarely on two of the surfaces this study says the engines cite most, not content stranded on a page the engine ignores.
The podcast finding is the sharpest tactical hook, because the citable artifact there is the transcript, not the talk. Feed Kompozy one episode and it fans that single input across its [output buckets](/glossary/output-buckets) — a blog recap and an email newsletter that render the best passages as crawlable, quotable text, clips that carry the on-the-record moments into YouTube and LinkedIn, and quote graphics for the social feed — turning a one-hour conversation into the written, indexable residue the companion study says decides whether a show gets cited at all. Then [Autopilot](/glossary/autopilot) handles the cadence, scheduling the approved set across the eight social platforms plus blog and email so the presence compounds week after week instead of depending on a single burst. The division stays honest: the earned surfaces are won by hand, and Kompozy manufactures, at volume, the crawlable off-domain footprint the ranking rewards. For the full strategy behind these three surfaces, see the guide on [Wikipedia, podcasts, and Reddit for AI citations](/guides/wikipedia-podcasts-reddit-ai-citations).
It is a 2026 index from Everything-PR, produced in coordination with 5W Public Relations, that synthesizes six independent citation studies — more than 680 million citations across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews, roughly from August 2024 to April 2026 — into one ranked list of the fifty domains AI answer engines cite most.
Across engines, Reddit ranks first, followed by Wikipedia, YouTube, LinkedIn, and Forbes, then Amazon, Business Insider, TechRadar, Reuters, and The New York Times. Reddit runs near 40% of multi-engine citation frequency, Wikipedia dominates inside ChatGPT, and the top fifteen domains together carry about 68% of all citations. Exact shares vary by study, engine, and query type.
Because the engine reads the transcript, not the audio. A companion 2026 podcast study found that shows failing to rank in AI citations overwhelmingly failed on transcript access and extractability rather than content quality. An episode with no accurate, indexed transcript is effectively invisible to an answer engine, so the written residue — transcript, show notes, recap, and captions — is the citable asset.
Only some of them. Wikipedia and Reddit are earned surfaces — self-editing a brand page gets reverted, and overt Reddit marketing gets removed — so you influence them through genuine, non-promotional work, not posting. But YouTube and LinkedIn, both in the top five, are platforms you publish to directly, which makes them the citation surfaces you can feed on purpose with consistent, expert-grade content.