The 84% myth: Why the PR industry’s AI numbers are wrong

David Clare is director and head of Fusion at Fire on the Hill. 

When an AI engine builds an answer, it draws on far more than press coverage. Earned media, the coverage PR teams strive for, is only one ingredient among several. So, when you hear that earned media drives more than 80%of AI citations, be skeptical. It is a comforting number, but is it too good to be true? (Spoiler: Yes, it is.)

At Fire on the Hill, we track how brands show up inside the likes of ChatGPT, Claude, Copilot, Gemini, Google AI Overviews, and many other LLMs every day, with more than 3 million data points in our own dashboards to date. From that wealth of data, we see an entirely objective view of what feeds LLM responses, and while earned media has a significant impact, the PR industry is overselling it and ignoring the other factors at play.

Is ‘earned media drives 84% of AI citations’ true?

The study making that claim is Muck Rack’s “What Is AI Reading?” report. Frequently cited by PRs, the report analyzed more than 25 million citations across 17 industries. We are fans of Muck Rack, and 25 million citations is a strong dataset to draw conclusions from, but the headline claim deserves scrutiny.

The study tested only three chatbots: ChatGPT, Claude, and Gemini. AI-search surfaces most people use, Google’s AI Overviews and AI Mode, were entirely left out. And its own data undercuts the headline: The single most-cited source on each platform is Wikipedia, PubMed, and Reddit, none of which any PR team earns through media relations.

The figure that maps to the work PR professionals do, journalism, is nearer a quarter, matching our findings. The problem is the 84% figure that is so often cited actually leans on a very generous definition of earned media: journalism bundled with academic research, government sources, Wikipedia, social platforms and third-party corporate content under a single label.

Speak to any client about earned media and they will be talking about media publications. And while other media beyond journalism can be earned, the way in which PRs are touting the 84% figure is clearly along the lines of “perform more media relations and win the GEO race.” This will only harm our industry in the long run. Brands will pour budget into media relations and fail to see the rewards they expect. Why? Because they’re missing out on three-quarters of what actually makes an impact.

Is search still bigger than AI?

While Muck Rack’s study completely left out search generative engines (i.e., Google), a recent YouGov report, “Searching for Answers: How AI is changing online discovery in 2026,” makes a different but equally misleading mistake.

YouGov surveyed 2,003 UK adults asking whether they used a “search engine” or an “AI assistant.” 85%said they’d used a search engine, versus just 31 percent who had used an AI assistant. The conclusion from this? Traditional search still dominates, and SEO matters more than GEO.

But there is a glaring flaw in such a conclusion. The report completely ignores the fact that search engines are AI themselves. Google made AI Mode the default search experience at its I/O conference on May 19, 2026, ten days before YouGov’s survey closed, and even stated “Google Search is AI search.”

Respondents may well be correct in answering that they use a search engine, but to draw the conclusion that this means GEO isn’t as important as SEO shows a rather astonishing lack of understanding of how modern search works. The fact is, those 85% are using AI. Either they or YouGov just don’t know it.

If we fight misinformation, why are we spreading it?

PR professionals counsel clients daily on accuracy, transparency, and evidence-based claims. Yet some of the most widely shared statistics in our own industry would fail the same scrutiny we apply to client communications.

If we want brands to trust our guidance on AI visibility, the data behind it needs to meet the standard we set for everyone else.