As artificial intelligence continues to reshape the landscape of digital discovery, the public relations industry finds itself at a critical crossroads regarding how it measures and reports on brand visibility. For decades, earned media—the coverage secured through traditional media relations—has been the primary currency of the PR professional. However, as Large Language Models (LLMs) such as ChatGPT, Claude, and Gemini become the primary interfaces for information retrieval, a significant discrepancy has emerged between industry claims and the technical reality of how these engines attribute information. A recent surge in optimistic data suggests that earned media accounts for the vast majority of AI citations, but a deeper dive into the data reveals a more complex and potentially troubling picture for brands relying on these metrics.
The Genesis of the 84% Statistic
The current debate centers on a widely circulated figure from the Muck Rack report titled "What Is AI Reading?", released in May 2026. The report, which analyzed a staggering dataset of 25 million citations across 17 different industries, concluded that a massive 84% of citations in AI responses were driven by "earned media." For PR agencies looking to justify budgets in an increasingly automated world, this number appeared to be a definitive proof of concept: traditional media relations were not only relevant but dominant in the age of AI.
The study focused on three primary chatbots: OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini. By examining how these models sourced their answers, the report aimed to provide a roadmap for "Generative Engine Optimization" (GEO). However, industry analysts, including David Clare, head of Fusion at Fire on the Hill, have begun to raise alarms about the methodology used to reach that 84% figure. The concern is that the definition of "earned media" used in the study is so broad that it obscures the actual impact of traditional journalism, leading to a potential misallocation of marketing and PR resources.
Deconstructing the Data: Journalism vs. General Information
The primary point of contention lies in the taxonomy of "earned media." In a traditional PR context, earned media refers to news coverage, feature stories, and editorial mentions in established journalistic publications. However, the 84% figure cited by the industry includes a massive "bundle" of diverse sources. When the data is disaggregated, a different story emerges.
According to internal data tracking performed by Fire on the Hill, which monitors over 3 million data points across various LLMs and Google’s AI Overviews, the actual percentage of citations attributed to traditional journalism is closer to 25%. The remaining 59% of the "earned media" category in the Muck Rack study is comprised of sources that PR teams rarely influence through standard media relations. These include:
- Wikipedia: Consistently the most-cited source across almost all LLMs due to its structured data and perceived neutrality.
- Reddit: A primary source for human-centric advice and reviews, bolstered by high-profile licensing deals between Reddit and major AI developers like Google and OpenAI.
- PubMed and Academic Repositories: Vital for technical, medical, and scientific queries.
- Government Portals (.gov): Used as the "ground truth" for regulatory and legal information.
- Third-Party Corporate Content: White papers and technical documentation hosted on non-journalistic platforms.
By labeling all these disparate sources as "earned media," the industry creates a false impression that increasing the volume of press releases and media pitches will result in an 84% dominance of the AI citation space. In reality, a brand that ignores its Wikipedia presence or its engagement on community platforms like Reddit is missing out on three-quarters of the visibility drivers.
The Search Engine Fallacy: SEO vs. GEO
The confusion within the PR and marketing sectors is further compounded by conflicting reports on how consumers are actually using these tools. A July 2026 report from YouGov, "Searching for Answers: How AI is changing online discovery in 2026," surveyed over 2,000 UK adults to determine the prevalence of AI assistants versus traditional search engines.
The findings suggested that traditional search remains the dominant force, with 85% of respondents stating they use a search engine, compared to only 31% who use an AI assistant. On the surface, this data suggests that Search Engine Optimization (SEO) should remain the priority over Generative Engine Optimization (GEO). However, technology analysts argue that this survey presents a false dichotomy.
The turning point occurred on May 19, 2026, during the Google I/O conference. Google officially transitioned its core product into "AI Mode," making AI Overviews the default experience for hundreds of millions of users. As Google executives stated at the time, "Google Search is AI search." Consequently, a user who claims they are using a "search engine" is, in most cases, interacting with a generative AI interface. The 85% of users identified by YouGov are already operating within the GEO ecosystem, whether they realize it or not. This lack of distinction in the data can lead brands to believe they are safe sticking to 2010-era SEO tactics, when the underlying technology has already shifted to a citation-based AI model.
Chronology of the AI Information Shift
To understand the current state of misinformation in PR data, it is helpful to look at the timeline of events that led to the May 2026 reporting cycle:
- Early 2024 – Mid 2025: Rapid adoption of LLMs leads to the "SGE" (Search Generative Experience) beta phase at Google. PR agencies begin to experiment with "AI-friendly" content.
- February 2026: OpenAI announces deeper integrations with news publishers, signaling a shift toward real-time data retrieval rather than just training on static datasets.
- May 19, 2026: Google I/O Conference. Google announces that AI-powered search is no longer an opt-in experiment but the standard global interface.
- May 29, 2026: The YouGov survey concludes, capturing a snapshot of a public that is using AI search tools but still labeling them as "traditional search."
- June 2026: The Muck Rack report gains viral traction in the PR industry, establishing the "84% earned media" narrative as a standard talking point for agency pitches.
The Risks of Strategic Misalignment
The danger of relying on inflated or misunderstood statistics is twofold: it threatens the credibility of the PR profession and it leads to strategic failure for clients. When PR professionals counsel clients based on the 84% figure, they are essentially promising that media relations is a silver bullet for AI visibility.
If a brand invests heavily in a media relations campaign and sees a surge in journalistic coverage but fails to see a corresponding rise in AI citations, the PR industry loses trust. The disconnect occurs because the AI engines are looking for "triangulation." An LLM is more likely to cite a brand if it finds a mention in a reputable news outlet (the 25%), corroborated by a Wikipedia entry, discussed on a Reddit thread, and supported by a technical paper in an academic database.
By focusing solely on the journalism slice of the pie, PR teams are leaving the majority of the "influence surface" to chance. A holistic GEO strategy requires a broader definition of "earned" that includes community management, technical SEO, and the strategic placement of information in open-source repositories.
Industry Reactions and the Call for Transparency
The critique of these numbers has sparked a debate among PR leaders. While some defend the Muck Rack report as a necessary tool for highlighting the value of the industry, others are calling for more rigorous data standards.
"We cannot fight misinformation for our clients while spreading it ourselves," says David Clare. The sentiment is echoed by digital strategists who argue that the PR industry must evolve its measurement frameworks. The consensus among data-focused PR practitioners is that "visibility" in 2026 must be measured through a multi-lens approach:
- Direct Citation Volume: How often is the brand mentioned by name in LLM responses?
- Source Diversity: Are the citations coming from a variety of domains (News, Wiki, Social, Gov)?
- Sentiment and Accuracy: Is the AI correctly interpreting the brand’s messaging?
- Conversion and Click-Through: For AI engines that provide links (like Google AI Overviews or Perplexity), is the traffic actually reaching the brand’s owned properties?
Broader Implications for the Future of Communications
As we move toward the latter half of 2026, the PR industry must reconcile its traditional strengths with the technical requirements of the AI era. The "84% myth" serves as a cautionary tale about the pitfalls of confirmation bias. While it is comforting to believe that the old ways of doing PR are more powerful than ever, the reality is that the digital ecosystem has become more fragmented and data-reliant.
For brands, the message is clear: media relations is a vital component of AI visibility, but it is not the whole. To truly "win" the GEO race, companies must ensure their data is accurate and present across the entire spectrum of sources that LLMs trust. This includes maintaining robust Wikipedia pages, engaging authentically in community forums, and ensuring that technical data is easily crawlable and "digestible" for AI agents.
The PR industry’s value proposition is shifting from "getting the story out" to "ensuring the story is found and cited." This requires a higher standard of data literacy and a commitment to transparency. If the industry wants to maintain its role as a trusted advisor to the world’s biggest brands, its own data must be as beyond reproach as the claims it makes for its clients. The era of the "comfortable number" is over; the era of objective, multi-source AI strategy has begun.
