Defining an ever-evolving term can be tricky.
Matt Dzugan is VP of Data & Intelligence and Linda Zebian is VP of Communications at Muck Rack.
We’ve seen some commentary recently about our “What Is AI Reading?” research, with some folks questioning whether our findings are skewed or misleading because of how we categorize earned media.
We’re thrilled people are talking about the data. We absolutely should be focused on where AI citations come from, how to get the right information into the right answers and how citation patterns are changing over time.
But much of the disagreement here is actually about the definition of earned media, something the PR industry has struggled to define for a long time.
Instead, we should be talking about what the data tells us as communicators and how we take action based on it, rather than getting hung up on vocabulary.
The data in question
We’ve now published three editions of “What Is AI Reading?”, in July 2025, December 2025 and May 2026. Across all three, journalism has accounted for 25% to 27% of links cited by ChatGPT, Claude and Gemini, a figure that has stayed remarkably consistent.
We’ve also found that the much broader pool of non-owned, non-paid sources has accounted for 82% to 89% of citations. In our research, we’ve called that broader category earned media. It includes journalism along with encyclopedic sources, government sites, academic sources, social platforms and other third-party content.
“Earned” is probably not a perfect label for all of those sources, and it’s always been a somewhat subjective term. Some practitioners use it almost exclusively to mean journalism, while others use it more broadly to describe attention or credibility a brand didn’t buy or publish itself.
That’s why we include the breakdown of the individual categories every time we publish the research. If you want to know how much comes from journalism specifically, you can see it clearly in the data.
The rationale for our definition is fairly simple. Our 2026 State of PR survey of more than 1,000 PR professionals shows just how much the communications remit has broadened. Content creation and influencer work are now among the top job functions, and the large majority of teams are active on LinkedIn and Instagram as core parts of their communications strategies. The job isn’t just press placements, and frankly, it never was. Our category structure reflects that.
The Wikipedia conundrum
One of the more specific questions raised is whether Wikipedia should really be considered earned media.
Wikipedia is the largest non-journalism category and the most-cited domain on ChatGPT in our research. We understand why some people wouldn’t put it in the same category as a New York Times article, but we also don’t think it makes sense to look at the two as completely separate.
To help explain why we bucketed it the way we did, we analyzed 489 S&P 500 company Wikipedia pages and the 58,490 references cited across them. Sixty percent of those references were journalism.
There’s a logical reason for that. Wikipedia has strict notability requirements, and for a company to warrant a page, it generally needs significant coverage from reliable, independent sources.
We’re also hearing from more PR teams that their company’s Wikipedia presence is becoming a GEO priority because of how often AI platforms cite it. But you can’t really separate a company’s Wikipedia presence from the body of journalism that exists about it. In many cases, that coverage is what makes the page possible and provides the sourcing behind what it says.
Looking only at the final URL cited in an AI answer can therefore give us an incomplete picture of where the information originated. A news story might be cited on Wikipedia, discussed on Reddit or referenced by another third-party site before some version of that information eventually appears in an AI answer.
None of that makes Wikipedia journalism, but it does make it a downstream source of journalistic coverage in many cases.
There probably won’t be one GEO number
Another point raised in the recent conversation is that GEO data can vary widely depending on what and how you measure. We agree.
ChatGPT and Gemini don’t always cite the same sources. Results vary by prompt, industry and time period. A citation also doesn’t necessarily mean a brand was recommended or even portrayed positively.
It’s worth understanding the scale of the research. Our latest study analyzed more than 25 million links cited by AI platforms, so comparisons to findings from much smaller samples need to account for differences in methodology and scale.
Either way, we shouldn’t expect one percentage to tell a company whether it has a strong AI presence. And we certainly wouldn’t tell a communications leader that our research means media relations drives 84% of AI visibility. It doesn’t, and our research has never said that it does. What it does show is that more than 80% of the sources AI platforms cite are sources brands don’t own or pay for.
We’re also not alone in finding that. AirOps analyzed 21,311 brand mentions and found that 85% came from third-party sources. The methodology and terminology are different, but the findings point in a similar direction: AI platforms rely heavily on information that comes from outside a brand’s own channels.
What matters for PR
For PR practitioners, the journalism number is significant on its own. Roughly one in four citations across three of the biggest AI platforms comes from journalism, and we’ve seen that percentage hold across three editions of the research.
But we’re equally interested in what happens beyond the direct citation. Reporting becomes a source for Wikipedia pages, gets discussed and linked to in online communities and can show up in other places that AI platforms pull from. We’re only beginning to understand how much of that secondary influence we can measure.
Media relations alone isn’t the answer to GEO, and PR can’t claim credit for everything that falls into that 84%. There are too many other sources involved and too much we still need to learn about how the platforms use them.
What we should pay attention to is how much AI pulls from sources outside a brand’s direct control, and the role communications can play in shaping the credible, independent information that exists about the organizations we represent.
We should keep questioning the numbers and comparing methodologies as more research comes out, including questioning ours. But whether we ultimately call that broader 84% category “earned media,” “third-party sources” or something else, the underlying findings remain the same. Paid content barely registers in what AI cites, journalism carries real weight, and the broader body of credible, third-party information about a brand is playing a major role in AI answers.
That’s what we want PR teams, and the people controlling their budgets, to act on.
The post What does ‘earned’ mean in the age of AI? appeared first on PR Daily.
