The rapid evolution of generative artificial intelligence has fundamentally altered the landscape of information dissemination, forcing a critical re-evaluation of how public relations professionals define and measure success. At the heart of this transformation is a burgeoning debate over the nature of "earned media" and its role in training and informing Large Language Models (LLMs). Recent research from Muck Rack, a leading public relations management platform, has ignited an industry-wide conversation regarding the sources that AI platforms like ChatGPT, Claude, and Gemini prioritize when generating responses. As the industry moves toward 2026, the traditional boundaries between journalism, third-party content, and brand-owned messaging are blurring, necessitating a more nuanced understanding of how credibility is established in a digital-first environment.
The Genesis of the Research and the Definition Debate
The discourse began following the release of the "What Is AI Reading?" research series, a multi-year longitudinal study conducted by Muck Rack’s data and intelligence teams. Led by Matt Dzugan, Vice President of Data and Intelligence, and Linda Zebian, Vice President of Communications, the study sought to identify the origins of citations provided by the world’s most prominent AI platforms. Across three distinct editions published between July 2025 and May 2026, the data revealed a consistent pattern: journalism remains a cornerstone of AI knowledge, accounting for approximately 25% to 27% of all citations.
However, controversy arose when Muck Rack categorized a broader pool of non-owned and non-paid sources—including Wikipedia, government websites, academic journals, and social media platforms—as "earned media." This broader category accounted for a staggering 82% to 89% of citations. Critics within the industry have argued that such a wide definition potentially misleads stakeholders by conflating traditional media relations with general third-party mentions. This tension highlights a long-standing struggle within the communications sector to define "earned" coverage in a way that satisfies both traditionalists and digital-age strategists.
A Chronology of the Muck Rack AI Research Series
To understand the current state of the debate, it is essential to trace the development of the "What Is AI Reading?" research. The study was launched to provide empirical data on Generative Engine Optimization (GEO), a new frontier for PR professionals seeking to ensure their brands are accurately represented in AI-generated answers.
- July 2025: The Inaugural Report. The first study established the baseline, showing that AI models were heavily reliant on high-authority third-party domains. It was here that the initial "84% earned media" figure began to circulate, prompting questions about the methodology behind grouping diverse sources.
- December 2025: The Mid-Term Update. The second report confirmed the stability of these figures. Despite updates to the underlying models of GPT-4 and Gemini, the ratio of journalism to other third-party sources remained largely unchanged, suggesting that AI training sets were prioritizing established credibility over real-time social buzz.
- May 2026: The Comprehensive Analysis. The latest edition expanded the sample size to over 25 million links. This massive data set reinforced the previous findings but also provided a deeper look into the "Wikipedia conundrum," revealing that the world’s largest encyclopedia serves as a vital bridge between journalism and AI answers.
The Wikipedia Conundrum: A Bridge Between Journalism and AI
One of the most contentious points in the current PR landscape is the classification of Wikipedia. As the most-cited domain on ChatGPT, Wikipedia holds immense power in shaping public and AI perception. Traditional PR metrics often exclude Wikipedia because it is not a "news outlet" in the classical sense. However, Muck Rack’s analysis of 489 S&P 500 company pages suggests that this exclusion may be a strategic error.
The analysis revealed that across nearly 58,490 references on these corporate Wikipedia pages, 60% were sourced directly from journalism. This creates a "downstream" effect: while an AI might cite a Wikipedia URL, the underlying information was originally "earned" through media relations and journalistic reporting. Because Wikipedia maintains strict notability requirements, a company generally cannot sustain a presence on the platform without a robust body of independent, reliable news coverage. Therefore, while Wikipedia is not journalism, it is frequently a secondary vehicle for journalistic impact.
Supporting Data and the Evolving PR Remit
The shift in how we categorize media is not merely academic; it reflects a fundamental change in the daily responsibilities of communications teams. According to the 2026 State of PR survey, which queried more than 1,000 professionals, the scope of the "communications remit" has broadened significantly.
Key findings from the survey include:
- Diversified Functions: Content creation and influencer management have moved into the top five job functions for PR teams, rivaling traditional media relations.
- Platform Priority: Over 75% of communications teams now identify LinkedIn and Instagram as core components of their strategy, treating these platforms not just as social outlets but as credible channels for brand authority.
- AI Integration: Nearly 90% of respondents reported using AI tools for drafting or research, yet only a fraction had a dedicated strategy for Generative Engine Optimization (GEO).
These statistics suggest that the industry is already practicing a broader version of "earned media" than it is willing to admit in its formal definitions. If a PR team successfully influences a Reddit discussion or a niche academic paper that eventually informs an AI response, that effort represents a form of earned credibility that falls outside the traditional "press placement" bucket.
Official Responses and Industry Reactions
The debate has prompted various responses from industry leaders and competing research firms. AirOps, a prominent AI search analysis firm, conducted its own study of over 21,000 brand mentions. Their findings largely mirrored Muck Rack’s, showing that 85% of brand mentions in AI search results originated from third-party sources.
While the terminology varies—some prefer "third-party credibility" over "earned media"—the consensus is shifting toward the idea that AI platforms are largely immune to direct brand influence through owned or paid channels. Paid content, in particular, barely registers in AI citations, accounting for less than 2% of the data analyzed. This has significant implications for marketing budgets, suggesting that investment in credible, independent third-party validation yields a much higher return in the AI era than traditional digital advertising.
Broader Impact and Strategic Implications for the Future
The implications of this research extend far beyond the PR department. As AI becomes the primary interface through which consumers and B2B buyers gather information, the "source of truth" matters more than ever.
1. The Rise of GEO (Generative Engine Optimization):
Just as SEO dominated the early 2000s, GEO is becoming the primary focus for digital strategy. However, unlike SEO, which relied heavily on technical tweaks and keywords, GEO is predicated on authority and "mention-graph" density. To be cited by an AI, a brand must be discussed by authoritative, independent sources. This elevates the importance of media relations, even if the final citation is not a direct link to a news site.
2. Credibility as Currency:
The fact that AI models favor journalism and academic sources suggests a "flight to quality." In an internet increasingly flooded with AI-generated "slop," the human-vetted nature of journalism provides a crucial signal of reliability for LLMs. PR professionals must therefore focus on high-impact, high-credibility placements rather than high-volume, low-quality distribution.
3. The Necessity of a Holistic View:
The debate over whether Wikipedia or Reddit counts as "earned media" may eventually become obsolete. For a brand, the goal is to occupy the "latent space" of an AI model—to be the entity that the AI associates with specific positive attributes. Achieving this requires a holistic approach that encompasses media relations, community engagement, and the curation of factual data on platforms like Wikipedia and Wikidata.
Conclusion: Moving Beyond Vocabulary
The current friction regarding the definition of earned media is a symptom of an industry in transition. While the label "earned" may be subjective, the data remains objective: AI platforms rely heavily on information that exists outside a brand’s direct control. For PR practitioners, the journalism numbers (25-27%) are a powerful testament to the enduring value of the fourth estate. Simultaneously, the broader 84% figure serves as a wake-up call that the ecosystem of influence is much larger than previously thought.
As the industry looks toward the late 2020s, the focus must shift from arguing over vocabulary to taking action based on how information actually flows. Whether it is called earned media or third-party validation, the objective remains the same: shaping the credible, independent narrative that defines an organization in the eyes of both humans and machines. The organizations that succeed will be those that understand the interconnectedness of the modern information ecosystem and invest in the long-term cultivation of authority across all third-party platforms.
