The 9 Publishers That Decide What AI Says About Your Brand

What percentage of LLM citations come from earned media?

The data is consistent across every major 2025–2026 study of AI citation behavior:

• 80% of LLM citations on branded queries come from earned media. (BuzzStream, 2026)
• 95%+ of AI citations come from organic, non-paid media. (FINN Partners, 2026)
• For branded queries, the breakdown is 48% earned media, 30% commercial content, 23% owned content (Omniscient Digital analysis of 23,000+ citations, 2026).
• Customer review queries: 82% earned media. (Omniscient Digital, 2026)
• General queries: 28% earned media, 22% owned content. (Penta Group via Axios, 2026)

Translation: when buyers ask ChatGPT, Gemini, or Perplexity about your category, the model is overwhelmingly synthesizing from third-party publications — not your blog.

Which publishers have AI licensing deals with OpenAI, Google, Meta, and Microsoft?

LLMs train on a finite list of publishers under contract. That list is now a closed asset class with a price tag attached to it:

• News Corp → OpenAI — $250M+ over 5 years
• Axel Springer → OpenAI — $25M+
• Reddit → Google — ~$60M/year
• Reuters → Meta — $25M one-time
• Taylor & Francis → Microsoft — $10M+ upfront, recurring through 2027
• Standard OpenAI offer to other publishers — $1M–$5M/year

Publishers with confirmed AI licensing deals include: AP, FT, Condé Nast, Hearst, Time, The Atlantic, Vox, The Guardian, Washington Post, Axios, USA Today Network, Schibsted, Le Monde — plus the five named above. (Sources: Press Gazette, Digiday, AI Watch.dog, 2025–2026.)

This is the privileged class. Their content is structurally weighted inside the models — both at training time and at retrieval time. Your homepage is not on that list. Neither is your blog. Neither is your podcast feed.

Why aren't websites enough for AI brand visibility?

LLMs train on a finite list of licensed publishers, not the open web. Your owned content competes against a closed asset class of contracted publishers — AP, Reuters, FT, Condé Nast, Hearst, and others — whose content is structurally privileged in AI training and retrieval. Coverage in those outlets is no longer just PR; it is the only durable input to AI brand visibility.

• AI summaries have driven a ~80% drop in publisher clickthroughs. (Futureweek, 2025) The traffic value of coverage is collapsing. The citation value is exploding.
• 82% of journalists now use at least one AI tool, up from 77%. (Muck Rack, 2026)
• 86% of journalists say PR pitches inspire at least some of their stories. (Muck Rack, 2026) Earned media is still the primary input to the content that feeds the models.

What PR teams are doing about it

• 76% of PR pros use generative AI (flat YoY — adoption has plateaued).
• 75% use at least one paid AI tool, up from 57%.
• 51% work at organizations with an AI use-case policy, up from 21% in 2024.
• Only 12% of PR pros currently use AI agents — the next frontier.
• 93% say AI helps them complete tasks faster; 82% say it improves quality.

Source: Muck Rack State of AI in PR 2026 (n=564).

The takeaway

LLMs cite a closed list of licensed publishers. Earned coverage in those outlets is no longer just PR — it's the only durable input to AI brand visibility.

Marketing teams optimizing schema and content alone are competing for the 5–22% of citation share that comes from owned content. The other 78–95% belongs to the publisher class — and the only way in is earned media. See the full SEO vs. AEO vs. GEO breakdown, or read how GEO is operationalized as a PR discipline at Crackle.