ChatGPT cited at least one source in 96% of its answers in Muck Rack’s May 2026 study, averaging five citations per response. Two years ago most chatbot answers cited nothing. Today the question for a brand is no longer whether AI assistants point to sources, but which sources they point to, how often yours is one of them, and how long that lasts.
This page collects the AI citation statistics that hold up: each number has a named source and a date. Where studies disagree, we say so. Methods vary a lot between them, so compare numbers within a study before comparing across studies. At the end is the five-line scorecard we use to turn these figures into something a marketing team can track each month.
How often AI answers cite sources

Citation is now the default in AI search, but the default differs by product.
In Muck Rack’s May 2026 “What Is AI Reading?” report, ChatGPT cited sources in 96% of responses, with an average of five citations each. That figure applies to the realistic consumer prompts Muck Rack tested, some naming brands and some not.
Google’s two AI surfaces behave differently from each other. Ahrefs compared AI Mode and AI Overview answers to the same queries using September 2025 US data and found that only 3% of AI Mode responses lacked a source, compared with 11% of AI Overviews.
When Google does show an AI Overview, it usually lists several sources. Pew Research Center’s 2025 analysis of real browsing data found that 88% of the AI summaries in its sample cited three or more sources and only 1% cited a single source. The median summary was 67 words long.
How often do those summaries appear at all? In Pew’s sample, 18% of Google searches produced an AI summary. Question-style searches triggered them far more often, and 36% of searches written as full sentences (with both a noun and a verb) produced one.
The takeaway is that for most commercial and informational queries, there is now a slot for a cited source above or beside the classic results. The fight is for those slots.
Which domains each platform prefers
Every major assistant has its own favorite sources, and the favorites shift.
Muck Rack’s May 2026 data shows the split clearly. ChatGPT’s single most cited domain was Wikipedia. Claude’s was PubMed Central, the free archive of biomedical research. Gemini’s was Reddit. Beyond Wikipedia, which all three share, the overlap was small. Among journalism outlets, only Forbes appeared in more than one provider’s top five, ranking first for Gemini and fifth for ChatGPT.
Reddit makes the point sharply. In that same study, Reddit accounted for about 2.4% of all citations, but almost all of it came from Gemini. ChatGPT cited Reddit 16 times. Claude cited it zero times.
Profound’s analysis of citations from August 2024 through June 2025 found a similar divide. Among ChatGPT’s top 10 cited sources, Wikipedia took 47.9% of the share, with Reddit at 11.3% and Forbes at 6.8%. Among Google AI Overviews’ top 10, Reddit led at 21.0%, followed by YouTube at 18.8%, Quora at 14.3% and LinkedIn at 13.0%. Perplexity leaned hardest on community content, with Reddit taking 46.7% of its top-10 share.
Pew’s data from Google’s AI summaries shows the same three giants at the top: Wikipedia, YouTube and Reddit together made up 15% of sources in AI summaries, close to their 17% share of links in standard results.
The practical meaning: a single “AI visibility” number hides a lot. A brand can be well cited by Gemini through Reddit threads and absent from Claude, which leans on research and established publications. Track each assistant separately.
How much journalism and earned media drive citations

The clearest finding across the Muck Rack series is that AI assistants cite coverage, not advertising.
In the May 2026 report, about 99% of cited links came from non-paid sources. Paid and advertorial content made up 0.3% of citations. About 84% of citations came from earned media, and journalism alone accounted for about 27%.
Press releases are a small and shrinking share of what gets cited directly. They made up 1.1% of citations in May 2026, down from 6% in Muck Rack’s December 2025 report and close to the 1.2% seen in July 2025. Muck Rack attributes most of the change to ChatGPT citing press releases less often. A release still matters, but mostly as the trigger for the coverage that gets cited.
The December 2025 report adds two useful details. Journalism held between 20% and 30% of citations at every point measured. And 51% of answers that cited Reddit also cited a journalistic source, which suggests the assistants pair community opinion with reported facts.
Freshness matters as well. More than half of all citations in the December 2025 study pointed to content published in the previous 12 months.
For a PR team, these are the most important AI citation statistics on this page. They say that a feature in a trade publication or a mention in a news story does double duty: human readers see it once, and AI assistants keep drawing on it for months.
One caution on reading these AI citation statistics as a ranking of outlets. A high share of citations for an outlet across thousands of prompts does not mean that outlet will be cited for your topic. Muck Rack found that top cited sources differ by industry, and a trade publication with a modest overall share can dominate the answers in its own niche. For most companies, the outlets that matter are the specialist ones their buyers already read.
How AI citations relate to Google rankings
Whether ranking in Google helps depends on which AI product you mean.
For Google’s own AI Overviews, rankings matter a great deal. Ahrefs found in July 2025 that 76.1% of pages cited in AI Overviews ranked in Google’s top 10 for the query. Another 9.5% ranked between positions 11 and 100, and 14.4% did not rank in the top 100 at all. The median ranking of a cited URL was position 3, and the first cited URL had a median ranking of position 2.
For other assistants, the link is much weaker. In an August 2025 study of 15,000 prompts, Ahrefs found that on average only 12% of the URLs cited by ChatGPT, Gemini and Copilot appeared in Google’s top 10 for the original prompt. Overlap with Bing’s top 10 was about 10%. Perplexity was the most aligned with traditional search results of the group.
Even Google’s two AI products disagree with each other. Ahrefs’ December 2025 comparison found AI Mode and AI Overviews cited the same URLs only 13% of the time for the same query, rising to 16% when comparing just the top three citations. Wikipedia appeared in 28.9% of AI Mode citations against 18.1% of AI Overview citations.
So classic SEO still earns AI Overview citations. It does far less for ChatGPT and Claude, which draw on a wider mix of sources that often do not rank for the exact query at all.
How fast citations change
AI citations are volatile. Any snapshot is out of date within weeks.
Ahrefs reported that 45.5% of AI Overview citations change when an AI Overview updates. Semrush tracked more than 100 million citations from 230,000 prompts across ChatGPT search, Google AI Mode and Perplexity over 13 weeks, from July 14 to October 12, 2025. In that window, ChatGPT cited Reddit in close to 60% of responses in early August, then dropped to about 10% by mid-September. Wikipedia saw a similar sudden fall on ChatGPT, even though both stayed among its most cited domains. On Google AI Mode, Wikipedia appeared in only about 2% of responses.
Muck Rack’s December 2025 report documented the same kind of swing: Wikipedia led ChatGPT citations in August before ChatGPT cut its reliance on the site, and YouTube briefly spiked in Gemini’s mix for a few weeks in November before falling back.
These swings are why a single audit is close to worthless. A brand that checks its AI visibility once a quarter is measuring noise.
Do citations send clicks?
Mostly no, at least not from Google’s AI summaries.
Pew Research Center tracked the browsing of 900 US adults in March 2025. When a search produced an AI summary, users clicked a link inside the summary on about 1% of visits. They clicked a traditional search result on 8% of visits with a summary, compared with 15% of visits without one. They were also more likely to end their browsing session entirely after a page with an AI summary (26%) than after one without (16%).
That changes what a citation is worth. It is less a traffic source and more a credibility signal: your name in the answer, attached to a source the reader trusts, at the exact moment they form a view. For brands, being the named answer can matter more than being the clicked link.
The Citation Share Scorecard
Statistics about everyone are only useful if you can turn them into numbers about you. We use a five-line tracker we call the Citation Share Scorecard. Run it monthly, on the same fixed set of 20 to 50 prompts that a buyer in your category would ask.
Line one is Citation Rate: the share of prompts where any page about your brand is cited. Line two is Earned Share: of those citations, the percentage that come from third-party coverage rather than your own site. The Muck Rack data suggests this should be the larger share. Line three is Platform Spread: your citation rate on each of ChatGPT, Gemini, Claude, Perplexity and Google AI Overviews, tracked separately. Line four is Freshness: the share of your cited pages published in the last 12 months. Line five is Accuracy: the share of answers that describe you correctly.
Your own AI citation statistics will move as much as the industry figures do, so log the date and the exact prompt wording each time, because the studies above show that citation mixes shift from week to week.
To sum up the numbers: AI assistants now cite sources in nearly every answer, they lean hard on earned media and journalism, each platform has its own favorite domains, Google rankings help mainly with AI Overviews, citation mixes change within weeks, and the click is rarer than the mention. Track your own share against that backdrop, and the statistics stop being trivia.