Earned Media is What AI Answer Engines Quote

Muck Rack's What Is AI Reading study, published in May 2026, analysed more than 25 million links cited by ChatGPT, Claude, and Gemini across 17 industries and found that 84% of those citations point to earned media: independent journalism and coverage a company did not write, pay for, or control. Paid and advertorial content accounted for 0.3% of the same set. Across all three editions of the study since July 2025, the earned-media share has held between 82% and 89%. A budget built mainly around owned content is optimising for the smaller slice of what these systems actually cite: the blog posts, landing pages, and press releases a company writes about itself.

What Counts as Earned Media in this Study

Muck Rack's distinction is the one PR teams already use day to day: earned media is coverage a third party chose to publish because it judged the story worth telling. It is different from content the subject paid for or wrote itself. Journalism specifically made up 27% of citations, holding within a narrow 25% to 27% band across every edition of the study. Owned assets, a company's own site and blog, and paid or syndicated content sit in what is left, and together they are a fraction of what earned media alone accounts for.

The Citation Behaviour Differs by Platform

The three models covered in the study do not behave the same way. ChatGPT cites a source in 96% of responses but averages only five citations each time, and Wikipedia is its most-cited domain. Gemini cites in 82% of responses with a wider net, averaging eight citations, and leans on Reddit as its top source. Claude is the most selective of the three, citing in 55% of its responses, but when it does cite, it pulls from an average of 13 sources, with PubMed Central leading. A campaign built around one platform's citation habits does not automatically transfer to the next one. Both the number of sources an answer draws from and the type of source it prefers shift by model.

Why Query Type Changes What Gets Cited

The study also found that what a query asks for changes what gets pulled in. Press releases appear almost exclusively in responses to industry-trend questions, at 3.5 times the rate they appear in best-of or recommendation queries. Independent reviews and comparisons dominate those best-of queries instead. A company issuing a press release about a product launch is writing for a different query type than a company hoping to be named in a best-of list, and the two need different source material behind them. Repurposing the same release for both is the wrong move.

What this Changes for a Marketing Budget

Most marketing budgets are still weighted toward owned content: the company blog, the landing pages, the case studies written in-house. None of that is wasted work. Owned content still does the job of converting a visitor once a person actually arrives. On this evidence, what it does not do is show up often as the source an AI system quotes when answering a question about the category. That role belongs mostly to journalism the company did not write: analyst coverage, trade press, and independent reviews that took the story on because it earned the space rather than because it was submitted. As this blog covered in the piece on what Google's own guidance says about AEO, GEO and the tools selling them, no third-party tool can see inside these systems either. An earned citation is the more reliable proof of visibility on offer, achieved rather than promised.

The practical shift treats media relations and analyst engagement as a citation strategy as much as a reputation one. A founder or a marketing lead pitching a trade publication, briefing an analyst, or securing an independent review is building the kind of source these systems already prefer to cite. That preference has held steady for a year across three separate editions of the same study.

Where to Start

Two moves make the shift concrete without a bigger budget. First, treat every piece of owned content as raw material for a pitch rather than the finished asset itself. The version a journalist or analyst writes independently is the one more likely to be cited later. Second, track which trade publications and analysts already cover the category, and prioritise relationships with them over a wider, shallower press list. The study's platform breakdown shows citation behaviour concentrating on a narrower set of trusted domains rather than spreading evenly across the web.

FAQs

What share of AI citations come from earned media? 84%, according to Muck Rack's May 2026 study of more than 25 million links cited by ChatGPT, Claude, and Gemini across 17 industries. The figure has held between 82% and 89% across three editions of the study since July 2025.

Do all three AI models cite sources the same way? No. ChatGPT cites in 96% of responses averaging five sources, Gemini cites in 82% of responses averaging eight, and Claude cites in only 55% of responses but averages 13 sources when it does.

Does this mean a company blog is pointless? No. Owned content still converts a visitor once they arrive. What the study shows is that owned content is cited far less often than earned media when an AI system chooses its sources, which argues for media relations alongside it rather than instead of it.

Should press releases be a bigger part of the strategy? Only for the right query type. Press releases appear almost exclusively in responses to industry-trend questions, at 3.5 times the rate they appear in best-of queries, so a release aimed at winning a comparison-style citation is aimed at the wrong outcome.

Weighing how much of your own content budget should move toward earned media? Reach out at hello@marcomfin.tech.