The rapid ascent of generative artificial intelligence (AI) has fundamentally altered the landscape of information retrieval, pushing the public relations and communications industry toward a critical inflection point regarding how "earned media" is defined and measured. As large language models (LLMs) like OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini become primary gateways for information, the traditional metrics used by communications professionals are being scrutinized under a new lens. Central to this evolution is the ongoing research conducted by Muck Rack, a leading public relations management platform, which has released a series of reports titled "What Is AI Reading?" This research aims to decode the DNA of AI-generated answers by analyzing the sources these platforms cite when responding to user queries.
The debate surrounding this data, led by Matt Dzugan, Muck Rack’s VP of Data & Intelligence, and Linda Zebian, VP of Communications, centers on a fundamental question: In an era where an AI might synthesize a dozen sources to answer a single prompt, what constitutes "earned" influence? While some industry purists argue for a narrow definition restricted to traditional journalism, the data suggests that AI platforms rely on a much broader ecosystem of third-party credibility, including encyclopedic entries, government databases, and academic repositories.
The Chronology of AI Citation Research
To understand the current state of AI search and its implications for PR, it is necessary to look at the timeline of Muck Rack’s investigation into generative engine behavior. The "What Is AI Reading?" project has evolved through three distinct phases, providing a longitudinal view of how AI citation patterns are stabilizing or shifting.
The first edition of the research was published in July 2025, marking one of the first large-scale attempts to quantify the sourcing habits of the major LLMs. This was followed by a second edition in December 2025, which began to show a trend of consistency in how models prioritized certain types of domains. The most recent update, released in May 2026, analyzed a massive dataset of more than 25 million links cited by AI platforms.
Across these three periods, a striking pattern emerged: journalism—defined as content from recognized news organizations—consistently accounted for 25% to 27% of all citations. Despite the rapid updates to the underlying AI models (such as the transition from GPT-4 to subsequent iterations), the reliance on journalistic integrity remained a constant pillar of the AI information diet. However, the broader category of "earned media"—which Muck Rack defines as any non-owned, non-paid third-party source—has maintained a dominant share of 82% to 89% of all citations.
Defining Earned Media in a Post-Search World
The tension within the PR industry often stems from the "PESO" model (Paid, Earned, Shared, Owned), which has served as the standard framework for media strategy for over a decade. Traditionally, "earned" media was synonymous with a news clip or a broadcast segment. However, Muck Rack’s 2026 State of PR survey, which queried over 1,000 professionals, reveals a significant broadening of the communications remit.
Modern PR teams are now tasked with content creation, influencer management, and community engagement on platforms like LinkedIn and Instagram. Consequently, the definition of "earned" has expanded to include any instance where a third party validates or discusses a brand without a direct financial transaction. This includes:
- Journalistic Coverage: News reports, features, and op-eds.
- Encyclopedic Sources: Wikipedia and specialized wikis.
- Academic and Government Data: Peer-reviewed journals and official .gov reports.
- Community Discussion: Reddit threads and forum entries.
- Social Proof: High-authority posts on professional networks.
Muck Rack’s researchers argue that focusing on whether a link is "strictly journalism" misses the forest for the trees. The critical takeaway for a brand is whether the AI is sourcing information from the brand’s own controlled channels (Owned) or from independent sources that provide third-party validation (Earned).
The Wikipedia Conundrum and Downstream Influence
One of the most debated findings in the Muck Rack research is the role of Wikipedia. In almost every study of AI citations, Wikipedia emerges as the single most-cited domain, particularly for ChatGPT. Critics have argued that categorizing Wikipedia as "earned media" is misleading, as it is a crowdsourced encyclopedia rather than a news outlet.
To address this, Muck Rack conducted a deep-dive analysis of 489 S&P 500 company Wikipedia pages, examining 58,490 individual references. The results provided a crucial link between traditional PR and AI visibility: 60% of the references cited on these Wikipedia pages were journalism.
This finding highlights the concept of "downstream influence." An AI may cite a Wikipedia article in its final answer to a user, but that Wikipedia article only exists—and only maintains its "notability" status—because of a robust foundation of journalistic coverage. Therefore, while a PR professional might see a Wikipedia link in an AI response, that link is often a proxy for the media relations work that secured the original news stories.
For communications teams, this means that the "Wikipedia presence" of a corporation has become a primary objective for Generative Engine Optimization (GEO). However, because Wikipedia has strict independent sourcing requirements, the only way to influence a Wikipedia entry is through the traditional "earned" route: securing high-quality, independent media coverage.
Generative Engine Optimization (GEO): The New Frontier
As traditional Search Engine Optimization (SEO) faces disruption from AI-powered "overviews," a new discipline is emerging: Generative Engine Optimization (GEO). Unlike SEO, which focuses on keywords and backlink profiles to climb Google’s rankings, GEO focuses on authority, citation frequency, and the sentiment of third-party mentions.
Muck Rack’s data is supported by external findings from firms like AirOps, which analyzed 21,311 brand mentions and found that 85% originated from third-party sources. This reinforces the reality that AI platforms are designed to be "objective" synthesizers. They are programmed to prioritize information that appears across multiple, credible, independent domains rather than relying on a brand’s own marketing copy.
The implications for PR strategy are profound. If 80% to 90% of what an AI says about a brand comes from sources the brand does not own, the value of media relations and third-party advocacy increases significantly. Conversely, "paid" content, such as sponsored posts or advertorials, barely registers in the citation data, suggesting that AI algorithms may be specifically tuned to discount or ignore content that is overtly promotional or lacks independent verification.
Industry Implications and Strategic Adjustments
The shift in how AI consumes information necessitates a change in how PR performance is reported to stakeholders. Communications leaders are increasingly moving away from "Ad Value Equivalency" (AVE) and toward metrics that reflect "AI Share of Voice" and "Citation Authority."
Key takeaways for PR practitioners based on the 2026 data include:
- Journalism is a Multiplier: Media relations remains the most effective way to seed the information ecosystem. A single high-authority news article can be cited by an AI directly, referenced in a Wikipedia entry, and discussed on social media, creating multiple "entry points" for the LLM to find and surface that information.
- The Authority of the .Gov and .Edu: Government and academic citations carry immense weight in AI responses. PR teams working in regulated industries or specialized tech sectors should prioritize placements and collaborations that result in mentions within these high-trust domains.
- The Decline of Owned Control: Brands must accept that they have less direct control over their narrative in the AI era. The "Owned" channel (the company website) is often treated as a secondary source of truth by AI models, which prefer to "verify" brand claims through third-party reporting.
- Methodology Matters: As more GEO data becomes available, practitioners must be wary of small sample sizes. Muck Rack’s analysis of 25 million links provides a macro-view that individual brand audits may not capture.
Analysis of the Future Landscape
The debate over the definition of earned media is more than just a squabble over vocabulary; it is a struggle to define the value of human-led communications in a machine-led information age. If the PR industry fails to claim credit for the broader ecosystem of third-party influence—including Wikipedia, academic citations, and community discussions—it risks undercounting its impact on the modern consumer’s journey.
As AI models become more sophisticated, they will likely get better at identifying "original reporting" versus "summarized content." This could potentially increase the citation percentage for journalism even further. However, the current reality remains clear: AI platforms are voracious consumers of the "earned" world. They seek out the credible, the independent, and the verified.
For the people controlling PR budgets, the data provides a compelling argument for continued investment in media relations and high-level strategic communications. While the tools for searching for information have changed, the fundamental requirement for brands to earn their reputation through independent validation has never been more critical. The 84% figure—representing the vast majority of AI citations—serves as a reminder that the world beyond a brand’s own website is where its true AI identity is formed.
