.png)
ChatGPT now cites sources in 96% of its responses, according to Muck Rack's latest analysis of AI citations. Ask it about your category and the answer arrives with receipts.
Those receipts point at what other people have written about you, which is why earned media just became the most important line in your marketing plan.
I've built a career on ideas people choose to talk about. Here's what earned media is, why AI engines lean on it so heavily, and how to earn the kind of coverage they cite.
What is Earned Media
Earned media covers any coverage, mention or conversation about your brand that you didn't pay for: press articles, reviews, social posts, podcast mentions and word of mouth.
Paid media is space you buy, and owned media is the channels you run. The earned share is the one other people choose to create, which is where its credibility comes from
Why earned media is now upstream of AI visibility
Third-party coverage now sits upstream of AI visibility. AI engines assemble their answers about a brand from what others have published, so the earned media you win today becomes the raw material for tomorrow's AI answers. While your own website still features in that supply chain, it’s now just one part of a larger ecosystem influencing citations.
The inputs in this ‘upstream model’ are earned coverage in the sources AI systems index, and the output is a footprint of references that engines draw on whenever they answer for your brand.
The use is measurement: AI citation share tells you whether that footprint is working.
Work from the inputs down: earn the coverage first, and the footprint and the measurement follow.
What AI engines actually cite
Earned media accounts for 84% of AI citations, according to Muck Rack's "What Is AI Reading?" study, which analyzed more than 25 million links cited by ChatGPT, Claude and Gemini. Paid and advertorial content made up just 0.3%. Journalism alone accounted for 27% of cited sources.
The finding is consistent. Across three editions of the study since July 2025, earned media has held between 82% and 89% of citations.
The pattern reaches beyond newsrooms. McKinsey's State of the Consumer 2026 report finds AI tools drawing on forums, retailer reviews, blogs and video platforms when they form answers about brands.
Put simply, the corpus AI reads is made of other people talking about you, and paid placement, at 0.3%, is pretty much just a rounding error within it.
That changes the brief for every marketer. Alongside rankings, track who's writing about you and where.
Earned coverage as a validation signal
Earned coverage works as a validation signal because a third party has put its own reputation behind the story. A journalist chose to write it, an editor chose to run it, and anyone can check it.
Trust has always worked this way offline. A friend's recommendation beats an ad because the friend has nothing to sell.
The Muck Rack numbers suggest AI engines have simply automated the friend. Greg Galant, Muck Rack's cofounder and CEO, called the findings "validation that the work they do to earn coverage in the right outlets has real consequences beyond traditional metrics."
Generative engine optimization (GEO) runs on the same logic. It rewards a reputation you earn from other people, which puts PR and creative ideas at the center of search. For founders, that puts the press list in the same conversation as the search budget.
An earned media example in practice: Stay in the Fight
In practice, earned coverage starts with an idea strong enough that journalists choose to cover it. In 2023, we made Stay in the Fight, a legacy campaign for Oxfam shot in London and Nairobi. We recast its older supporters, usually addressed as donors, as the original protest generation.
The film follows Edith, a retired activist, as she tells her grandchildren how a gift in her will keeps her in the fight. It cuts between her 1980s London protests and climate activism in Kenya today. It was also Oxfam's first TV ad in more than a decade.
We also gave journalists a news hook. A full page in the Observer marked the 40th anniversary of the 1983 Hyde Park peace march, with original protesters reunited for the photography.
The idea did the distribution work. The campaign earned more than 240 pieces of coverage, including the Evening Standard, The Independent and Mail Online, with a combined reach of more than 50 million, according to the SOFII case study.
During the October to November 2023 TV window, legacy consideration among the target audience rose 10%, and new inquiries and pledges rose 28% year on year. Every one of those articles is a source an AI engine can and will read when it wants to know more about our work.
How to measure earned media in the AI-search era
You measure earned media by how much coverage you win, where it lands and what it moves. The established measures are coverage volume, share of voice and earned media value. In the AI-search era, add a fourth: AI citation share, meaning how often AI engines cite sources that mention your brand.
Coverage volume counts the pieces. Share of voice compares your mentions with competitors' over the same period, which makes it more honest than a raw count.
Earned media value puts a dollar figure on coverage, usually by pricing it as if it were bought advertising. It's handy for a board slide and weak as a measure of outcomes, because it rates a mention by what the space would have cost, not by what it changed.
AI citation share closes the loop. Track a fixed set of prompts your buyers would ask, then record which sources the engines cite and whether your coverage is among them.
Set a baseline before a campaign launches and rerun it monthly across several engines, because each one cites differently. Muck Rack found that Claude cites sources in 55% of responses but averages 13 citations when it does, while ChatGPT averages five.
How to win earned media that AI engines cite
Coverage that AI engines cite rests on three things: an idea worth covering, placement in the sources AI actually reads, and a story that's easy to quote. The idea decides whether the other two get a chance. Get all three right and the coverage keeps working long after the news cycle moves on.
Start with an idea worth covering
Coverage starts with a story; no story, no coverage. Journalists pick what will surprise their readers, so the idea has to clear that bar before any press release or distribution list comes into play.
The ideas that travel carry tension: they say something true that people have only been thinking, or show a familiar truth from a surprising angle. Before you plan a single pitch, ask whether a reporter would want this story with your logo taken off it. Rework the idea until the answer is yes, then open the press list.
Stay in the Fight earned its coverage by flipping an assumption about older people. The idea carried it, and that's the bar.
Earn coverage in the sources AI reads
Some coverage counts for far more than the rest. AI engines lean on authoritative, widely indexed publications, and Muck Rack found a single news outlet, Axios, among ChatGPT's top three cited domains in 13 of 17 industries.
A piece in a respected national or trade title keeps surfacing in AI answers long after a burst of vanity pickup has vanished. Map the titles your buyers read and AI engines cite, then build lasting relationships with the journalists who write for them.
Trade titles earn a place on that map. They're specific, well indexed and read by exactly the people making buying decisions. Forums and review sites count too, so show up in the communities where your customers already compare notes.
Make the story easy to quote and repeat
Give reporters, and the AI engines that read them, one clear claim or figure to lift. If your story needs three paragraphs of context before it makes sense, it won't survive being summarized. Put the number, the finding or the point of view in one sentence a stranger could repeat.
Then say it the same way everywhere: in the press release, in your spokesperson's quotes and on your own site. Consistent repetition across independent sources is how a claim becomes the answer.
Original data is the easiest thing to quote. A survey, an index or a finding of your own gives journalists a reason to cite you, and gives AI engines a fact to attach your name to.
Frequently asked questions
Is earned media the same as PR?
PR and earned media overlap, and PR is the discipline behind much of it: pitching journalists, managing reputation and shaping narratives. Earned media is one result of that work, and it also comes from customers, creators and communities who talk about you entirely on their own.
What is an example of earned media?
Earned media examples include a newspaper article about your campaign, an independent product review, a podcast host recommending your brand and a customer's post that spreads on its own. Each one shares a trait: someone outside your company chose to create it and decided what it said.
What is earned versus unearned media?
The earned versus unearned media split is about who chose to put your brand in front of people. Unearned media covers paid placements, such as ads and sponsored content, plus the owned channels you publish on directly. Unearned media buys reach on your schedule, while earned media borrows credibility on someone else's.
How do you optimize earned media for GEO?
To optimize earned media for GEO, make your coverage easy for AI engines to find and attribute. Make sure articles name your brand in full, link to a live page on your site and sit on open, indexable pages that AI crawlers can reach. Then pitch follow-ups to titles that have already covered you.
Put earned media to work in AI search
Earned media has always made campaigns famous. Today it also supplies the material AI engines use to decide what your brand is, whether you're worth recommending and who else belongs in the answer. The brands that show up in tomorrow's answers are the ones people are writing about today.
Don't Panic’s work is built around this approach: ideas designed to generate discussion and earn attention in the places that matter. As an earned media agency, Don't Panic puts the idea at the centre of the process, recognising that the ideas most likely to earn coverage are also the ones most likely to be picked up and referenced by AI systems.
