Earned media in 2015 was a finite resource. A brand pursued press hits the way a sales team pursued accounts — one journalist, one outlet, one story at a time. The hit landed, the brand harvested traffic and credibility for the news cycle, and the cycle closed.
Earned media in 2026 still works the same way at the human-attention layer. A second layer has been added underneath: every earned media hit now feeds the AI engine corpus that ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews retrieve from when answering questions about the brand. The hit no longer ends with the news cycle. It compounds in the engine cycle for years. AI Communications as a discipline measures both layers.
What changed about earned media value
The 5W AI Visibility Index research across multiple verticals documented that AI engines weight earned media citations differently than search engines did. Trade-press citations carry more weight than general-interest hits. Primary-source publication coverage carries more weight than syndicated coverage. Coverage that names specific outcomes, specific dates, and specific brand attributes gets retrieved. Coverage that mentions the brand in passing does not.
The earned media playbook from 2015 — reach journalists, provide useful information, do something newsworthy — still works as the foundation. The 2026 layer adds: optimize each hit for engine retrieval, not just human attention.
Three updated rules for earned media in the AI engine era
One — reach the journalists whose work compounds. Not every outlet feeds the engine corpus equally. The collapse of press-coverage-as-AI-authority documented that mid-tier coverage often fails to enter the corpora that drive retrieval. Concentrate effort on outlets whose work consistently gets cited by the engines.
Two — provide useful information that the engines can extract. Numbers, named outcomes, structured data, specific dates. The engines retrieve specific facts more reliably than vague claims. A pitch that gives the journalist three named facts the engines can extract creates an earned media hit that compounds in retrieval. A pitch with adjectives produces a hit that fades.
Three — do something the engines will treat as primary source. Original research the brand owns. Indices the brand publishes. Founder commentary on a verifiable event. The engines treat these as primary sources and retrieve them across multiple queries for years. Stunts, sponsorships, and announcements without primary-source substance get rendered into the news cycle and forgotten by the engines.
The earned media metric stack in 2026
News-cycle metrics still matter — coverage volume, reach, sentiment, share of voice. Engine-cycle metrics matter more — Citation Share across the engines, retrieval frequency per topic, engine-rendered sentiment, named-entity disambiguation.
The brand measuring only news-cycle earned media in 2026 is operating with half the dashboard. The 5W research program measures both layers across every category. Everything-PR coverage tracks which outlets compound in the engines and which fade.
Earned media still requires finesse and the right operator. The definition of "the right operator" updated. The engines are the audience the brand earns the media for.
Originally published October 2015. Cleaned up and republished June 2026.
Ronn Torossian is the founder and chairman of 5W AI Communications, the AI Communications Firm. He is the publisher of Everything-PR and the author of two best-selling editions of For Immediate Release.