Your SEO Ranking No Longer Predicts AI Citations, New Research Shows

A 5.35 million-citation study found earned media drives 84% of AI answers while paid content barely registers, and what that means for AEO.

Your SEO Ranking No Longer Predicts AI Citations, New Research Shows
Quick answer

Kingsworth Research analyzed 5.35 million AI citations across ChatGPT, Claude, Gemini, Perplexity and Grok and found that earned media (journalism and independent third-party coverage) drives about 84% of AI citations, while paid and advertorial content accounts for roughly 0.3%. Ranking well on Google no longer predicts whether you get cited in an AI answer.

Key takeaways
  • Non-paid sources make up 94-95% of all links AI engines cite, across five major engines.
  • Journalism specifically accounts for about 27% of citations on its own, more than any other single category.
  • Paid and advertorial content is nearly invisible to AI answer engines, at roughly 0.3% of citations.
  • Your own site ranking well in Google search is not the same signal as being cited in an AI answer.
  • The practical shift: getting covered by real, independent publications now matters as much as your own on-page work.

What the study actually found

Kingsworth Research, an independent firm covering generative engine optimization, published the analysis this month. The headline finding: traditional search rankings have decoupled from AI citation share. A page ranking on page one of Google carries no guarantee it will ever be quoted by ChatGPT, Perplexity, or Google's own AI Overviews.

Share of AI citations by source type
Earned media 84% Owned content 15% Paid / advertorial 0.3%
Source: Kingsworth Research, 5.35M-citation analysis across ChatGPT, Claude, Gemini, Perplexity and Grok.

The breakdown is stark. Earned media, meaning journalism and independent coverage nobody paid for, drives about 84% of citations across the five engines studied. Paid and advertorial content, the kind you'd buy placement for, comes in at roughly 0.3%. Owned content (your own site) fills most of the remaining share.

Why this lines up with how these models actually work

This isn't a surprising result if you think about what a generative model is actually doing when it answers a question. It's not ranking pages by a formula the way a search index does, it's trying to synthesize a trustworthy answer, and it weighs a claim differently depending on who's making it. A claim on your own product page reads as self-interested. The same claim reported by a journalist, or corroborated across several independent sources, reads as verified.

Traditional SEO rankings are no longer a reliable proxy for AI visibility.

That's the study's own framing, and it matches something we've written about before: the Princeton GEO research found that citing named, authoritative sources was one of the two strongest levers for AI citation, right alongside using concrete numbers instead of vague claims. This new data is the same finding at a bigger scale, applied to where the citations themselves come from, not just how a single page is written.

What to actually do about it

None of this means on-page AEO work stops mattering. It means it's necessary but not sufficient. A few concrete moves:

  1. Get covered, don't just publish. If you have something genuinely newsworthy (a real data point, a real product change), pitch it to a journalist or an independent publication instead of only writing it up yourself.
  2. Make your own pages easy to quote when they are cited. Named sources, concrete numbers, and clean structured data still decide whether a model can lift a fact out of your page cleanly. Run a full AEO & GEO audit to see exactly where your own pages fall short on that.
  3. Check what's actually reachable first. None of the earned-media work matters if the AI crawlers can't read your site at all. A free LLM Access Check shows you that in about 20 seconds, before you invest in anything else.
  4. Track who's citing you, not just who's linking to you. Backlinks were always a proxy for authority. Citations in AI answers are a more direct read on whether that authority is actually being recognized right now.

The uncomfortable part for smaller sites

If earned media is 84% of the citation pool, then a small site with no press coverage and no independent mentions is starting from close to zero, no matter how well-optimized its own pages are. That's a harder problem than fixing a schema tag. It also means AEO and PR, which used to live in completely separate parts of a marketing org, are now solving the same problem from two different angles.

Citation source Share of AI citations
Earned media (journalism, independent coverage) ~84%
Owned content (your own site) ~15%
Paid / advertorial ~0.3%

The table above uses the study's own reported categories. The exact percentages will shift as more research like this comes out, but the direction is consistent with everything else we've seen: AI answer engines trust what other people say about you more than what you say about yourself.

FAQ

Does this mean SEO is dead for AI visibility?

No, but it means SEO ranking alone doesn't transfer to AI citation the way it used to. Technical AEO fundamentals (crawlability, structured data, clear answers) still matter, they're just not the whole picture anymore.

How do I get earned media coverage if I'm a small site?

Start with something genuinely specific and newsworthy, a real data point from your own audience or product, not a generic opinion piece, and pitch it directly to journalists or publications who cover your niche.

Is this study peer-reviewed research?

No, it's industry research from Kingsworth Research, a firm covering generative engine optimization, published as their own analysis rather than an academic paper. Treat the direction of the finding as more solid than the exact percentages.