AI search visibility is not won by publishing more blog posts or buying more placements alone. A new October edition of Muck Rack and Generative Pulse’s What Is AI Reading? study points to a broader source ecosystem: more than 84% of AI citations came from earned and other third-party sources, while the share going to journalism fell and owned media gained ground. For business leaders, the signal is not “stop publishing” or “do more PR.” It is to build credible evidence across both your own site and the places AI systems already trust.

The fourth edition draws on Generative Pulse data covering more than 40 million AI responses and 207 million citations from January through September 2026, plus a September test of 1,020 unbranded consumer prompts across ChatGPT, Claude, Gemini and Perplexity. The study is vendor-produced and its public summary does not provide confidence intervals or full prompt-level sampling details, so treat it as directional evidence, not a guarantee that any single tactic will lift rankings, citations or sales. Muck Rack’s report and methodology summary and its October 28 release provide the underlying figures.

The headline is stable; the source mix is changing

Across four editions, Muck Rack says earned and third-party sources have supplied between 82% and 89% of citations. In the latest cut, about 99% of cited links were non-paid, while paid and advertorial content accounted for just 0.3%. That does not mean paid media is useless for demand generation. It means paid placements are not a substitute for the independent evidence AI answers often reference.

The notable change is inside that broad pool. Journalism’s share fell from 27% in May to 22% in September. Government and nonprofit sources, encyclopedic sites, academic research and third-party corporate content all rose; owned media also increased from 13.7% to 15.2%. Muck Rack CEO Greg Galant said, “AI’s sources are changing, but communicators are still the ones best positioned to lead on GEO because the work of earning credibility with third parties is exactly what AI rewards.” The practical reading is balanced: credible coverage remains valuable, but a press-only strategy misses other source types and a website-only strategy misses the larger third-party ecosystem.

“Earned” is broader than journalism

One important nuance: the report’s greater-than-84% figure includes more than news articles. It groups journalism with government and NGO websites, academic research, encyclopedic sources and third-party corporate content. Those sources play different roles in a buyer’s decision, and the aggregate statistic should not be translated into “84% of AI citations are press coverage.” Muck Rack’s own breakdown puts journalism at 22%, while third-party corporate content reached 25.1% and owned media 15.2%.

That distinction changes the operating plan. A business selling specialized services may need verifiable case studies, professional-association references, detailed product documentation and independent reviews alongside media outreach. A company with original data can make it easy for journalists, analysts and industry sites to cite the underlying evidence. The goal is not to manufacture mentions; it is to make accurate, useful information available wherever buyers and credible intermediaries look for it.

Build for the question and the platform

The same study found that source preferences vary by query. Practical guides supplied 68% of citations for how-to questions, buyer’s guides and reviews appeared in 45% of “best-of” answers, and expert analysis led 41% of industry-trend answers, according to Axios’ report on the study. Platform differences matter too: ChatGPT leaned most on government and nonprofit sources in the sample, Gemini on company-owned content, Perplexity on journalism, and Claude on third-party corporate content.

These are not universal ranking rules. The report also found that when the same prompt was run twice on the same day, only 41% of ChatGPT’s cited domains and 45% of Gemini’s repeated, versus 94% for Perplexity. A one-off prompt check is therefore a poor scorecard. Measure a stable set of buyer questions repeatedly, across several platforms, and track whether your brand is mentioned, which sources support the answer, and whether the answer is accurate.

What businesses should do now

  • Map questions to evidence. List the how-to, comparison and industry questions buyers ask, then identify what proof each answer needs: instructions, specifications, reviews, expert commentary or independent data.
  • Upgrade owned assets. Make product and service pages precise, current and easy to verify. Add useful guides, transparent methodology, named expertise and clear supporting evidence rather than generic keyword copy.
  • Earn third-party validation. Prioritize relevant trade publications, credible reviews, associations, research partners and expert sources. Evaluate coverage for audience and authority, not just link counts.
  • Measure across platforms. Re-run consistent prompts on a schedule, record mentions and cited URLs, and compare changes with referral traffic, qualified leads and branded demand. Treat citation shifts as a diagnostic, not proof of causation.

AI search is making reputation, content quality and discoverability parts of the same growth system. Real Internet Sales helps businesses connect SEO, GEO and digital marketing into a practical plan. Call 803-708-5514 or visit realinternetsales.com to discuss how your brand can be easier to find, verify and trust.