Earned Media Drives 84% of AI Citations. Owned Media Is What Makes It Possible.

A Muck Rack study has been getting passed around PR teams for weeks, and most people are taking the wrong lesson from it.

The study, Muck Rack's "What Is AI Reading?", looked at more than a million links cited by ChatGPT, Claude, and Gemini. Muck Rack's headline finding was that earned media accounts for roughly 84% of the sources these models cite, while paid content sits near zero. Cue the victory laps. AI loves journalism, so buy more PR.

I'd hold off on that conclusion.

The number is real. But the lesson people are pulling from it skips the more useful half of Muck Rack's research, which is about which earned media the models trust and when they bother to cite it. Read that part, and the case for dumping everything into press placements gets a lot weaker. The case for owned media gets a lot stronger.

What the models actually reward

Muck Rack found the two biggest models have very different taste.

ChatGPT leans toward large, mainstream publications, the wire-service and national-brand tier with enormous reach. Claude goes the other way. According to Muck Rack's breakdown of Claude's citations, it pulls heavily from niche outlets, trade publications, and specialized sources, and its hundred most-cited outlets average roughly half the monthly traffic of ChatGPT's. Claude isn't impressed by audience size. It rewards outlets that clearly own their subject.

The timing gap is just as sharp. Both models prefer recent content, but Muck Rack found they define recent differently. ChatGPT front-loads everything: a story gets cited hard in its first week, then the citations fall off fast. Claude keeps citing the same coverage for weeks, with a window Muck Rack pegs at roughly the past ten. One model pays out for a single splash. The other pays out for showing up week after week.

That second pattern is the one worth obsessing over.

You can't rent a rhythm

If consistency over weeks is what gets rewarded, here's the uncomfortable math for anyone betting the strategy on PR. You don't control consistency in earned media. You control almost none of it.

No trade editor is going to run your story on the schedule your visibility needs. You can't promise yourself a placement in week three to keep the momentum from week one alive. You can't manufacture a steady ten-week drumbeat of coverage, because the timing and the angle and the yes all belong to somebody else. That's the nature of earned media. You're borrowing another outlet's credibility, on their calendar, when they feel like lending it.

A one-time splash, you can sometimes engineer. A rhythm, you can't. And the rhythm is what Muck Rack's data says the models score you on.

Owned media is the only channel where you set the calendar. Your blog, your newsletter, your research, your published point of view on a domain you control. You pick the topic, the depth, the publish date. You ship in week one, again in week three, again in week six, on purpose. The behavior the study rewards is the behavior only owned media lets you produce on command.

Depth is something you build, not something you ask for

Go back to what Claude favors: niche and trade sources with real authority on a topic. Muck Rack frames this as topical authority beating raw reach. A small outlet that is obviously about a subject outranks a giant one that mentions it once.

Strip the media-targeting language off that and you've described what owned media is supposed to do. A company that publishes credibly and often on its own area of expertise builds topical authority directly, in a format it owns, without waiting for a reporter to bless it first. That's the signal these models hunt for, and you can produce it yourself.

There's a second reason owned media matters here, and PR-first thinking misses it completely. Earned coverage doesn't come from nowhere. A reporter needs an angle, a number, a quote, a take worth arguing with. More often than not, that raw material starts as owned media. Your original research turns into the stat a trade outlet runs. Your founder's post turns into the perspective a journalist calls to expand. Owned content is what gets you covered in the first place.

So even if all you care about is earned citations, you still need an owned media operation feeding the machine.

Where this leaves you

Earned media still counts. Muck Rack's 84% figure is not a rounding error, and anyone telling you to fire your PR team is misreading the study as badly as the people treating it as a press-release jackpot.

The point is the order of operations. Build the owned media engine first, because it's the only part you actually own, it produces the cadence and the depth the models reward, and it generates the material that earns you coverage. Then use earned media to amplify it.

Owned media spent years as the first thing companies cut when search traffic went soft. That made sense when the whole game was ranking on a results page you didn't control. The game moved. More and more, the first answer people see is generated by a model, assembled from sources picked for consistency, depth, and relevance over time. That's the ground owned media was built to win on.

The study everyone's quoting to justify a bigger PR budget is making the case for the one asset you fully control.