Earned vs. Owned AI Citation Rates: What the 2026 Data Actually Shows
Muck Rack's analysis of 25 million+ AI-generated links puts sources brands neither own nor pay for at 84% of AI citations, 82-89% across three editions. Narrower measures read lower. Here is what each number counts and how to allocate against it.
Sources brands neither own nor pay for account for 84% of AI citations across ChatGPT, Gemini, and Claude, and have held between 82% and 89% across three editions of the study. Brand-owned websites generate single-digit citation rates. This is not one study saying this. It is five independent analyses of more than 25 million links arriving at the same conclusion.
Before you spend against that number, know what it counts. The published figures for "earned media's share of AI citations" range from 37.6% to 85%, and all of them are correct — they count different things. Which one your budget should answer to depends on which category a placement you can actually buy lands in, and that is the second section below.
I have spent nearly a decade placing brands in earned media. What changed is not the channel. What changed is who is reading. Before a human ever sees a search result, an AI engine has already decided which sources to cite. And those engines are choosing earned media over brand-owned content at a ratio that should make every founder rethink where their marketing dollars go.
Five Studies, One Conclusion
The convergence across independent research is what makes this finding load-bearing, not any single data point.
Muck Rack's Generative Pulse study analyzed more than 25 million links from ChatGPT, Claude, and Gemini responses across 17 industries. The result: 84% of all cited links came from sources brands neither own nor pay for. Paid and advertorial content together accounted for 0.3% of cited links in Muck Rack's May 2026 sample. That 84% is a wider category than press coverage — it bundles journalism with Wikipedia, Reddit, PubMed, government and academic sources — and journalism alone makes up 27% of cited sources, spanning more than 20,000 distinct outlets. Across three editions of the study from July 2025 through May 2026, earned media has ranged between 82% and 89%. The number barely moves. The measured unit is source composition inside Muck Rack's observed citation sample. Boundary to preserve: Muck Rack Generative Pulse does not establish that earned media causes citation, that any one placement will be cited, or that earned media is a universal engine-selection mechanism.
5W's research on AI brand discovery tested the same question across six AI engines and found that the sources appearing in AI answers bear almost no relationship to the pages ranking first on Google for the same queries.
Arfadia's aggregated citation statistics put the figure at approximately 82%, with an additional finding: including statistics in content boosts AI visibility by 41%.
Omniscient Digital's analysis of 23,387 citations found an even starker gap for branded queries specifically: 68 to 85% of citations come from third-party sources rather than the brand's own website. When someone asks an AI engine about your company, the AI is more likely to cite Forbes or TechCrunch writing about you than your own about page.
Foglift Research's Q2 2026 benchmark corroborates the pattern, and the GEO study (Aggarwal et al., KDD 2024) showed GEO-optimized content achieves up to 40% higher visibility in AI-generated responses.
The data is not ambiguous. Earned media is not slightly better for AI citations. It is the dominant channel by a factor of 5 to 10.
What the 84% Counts, and Why Another Study Says 37.6%
If you have seen a much lower number quoted for the same claim, it is not a competing estimate. It is a narrower category.
Muck Rack's 84% is everything brands neither own nor pay for, measured across three engines. That bucket is journalism plus Wikipedia, Reddit, PubMed, government and academic sources. Journalism on its own is 27% of cited sources in the same sample.
Meltwater's GenAI Lens measures eight engines and splits the same web into six categories. Its "Earned/News" line — press coverage and industry publications only — read 3,006,799 citations, or 37.6%, in May 2026, down from 39.5% in April. Its 50.3% "Other" bucket is not an owned-content figure either: it bundles corporate sites and blogs with the long tail of third-party domains.
Both numbers are accurate. Neither one answers "earned versus owned" cleanly, because no published study draws that line where a buyer draws it. The two figures that are unambiguous:
| Figure | What it counts | Source |
|---|---|---|
| 84% (82–89% over three editions) | Not brand-owned and not paid: journalism, encyclopedias, community, academic, government | Muck Rack, May 2026, 3 engines |
| 37.6% (39.5% in April) | Press coverage and industry publications only | Meltwater GenAI Lens, May 2026, 8 engines |
| 27% (25–27% over three editions) | Journalism alone | Muck Rack, May 2026 |
| 0.3% | Paid and advertorial content | Muck Rack, May 2026 |
| 0.2% (0.4% in April) | Press releases | Meltwater GenAI Lens, May 2026 |
The allocation question answers to the 27% and the 37.6%, not the 84%, because those are the only lines a PR budget can buy into. You cannot purchase a Wikipedia citation or a Reddit thread. What you can buy is editorial coverage, and that is the slice that slipped from 39.5% to 37.6% between April and May — while press releases, the thing most budgets actually fund, halved from 0.4% to 0.2%.
A source-by-source synthesis of the independent studies behind these figures — including which domains hold that share per engine — is published on Machine Relations, which maintains the measurement side of this question.
Why AI Engines Prefer Third-Party Sources
The mechanism is structural, not accidental.
AI citation systems are built to assess source credibility. A claim published on your own website carries inherent bias. The same claim validated by an independent publication carries corroboration. This is exactly how academic citation works, and it is exactly how LLMs have been trained to evaluate source authority.
When ChatGPT decides whether to cite a stat about your company, it is weighing the same question a journal reviewer would: does this source have a reason to inflate the claim? A third-party publication that verified your data independently passes that test. Your marketing page does not.
This is why the Muck Rack data shows journalism at 27% of all cited sources. Journalism carries editorial verification. It carries institutional credibility that AI systems can measure through training data patterns, cross-referencing frequency, and domain authority signals. Reddit (Gemini's top cited domain) carries community validation. PubMed Central (Claude's top cited domain) carries peer review.
Brand blogs carry neither.
The Platform-by-Platform Breakdown
Not all AI engines cite the same way, and the differences matter for where you invest.
ChatGPT cites sources in 96% of its responses, averaging 5 citations per response. Its top domain is Wikipedia. It is the most citation-dense engine and the one most likely to surface earned media from traditional journalism outlets. Axios appears in ChatGPT's top 3 domains across 13 of 17 industries studied. No other journalism outlet achieves that breadth.
Gemini cites in 82% of responses with an average of 8 citations per response. Its top domain is Reddit. Community-validated, discussion-format content performs disproportionately well here.
Claude cites in 55% of responses but averages 13 citations when it does cite. Its top domain is PubMed Central. Claude skews toward academic, research-grade, and deeply sourced content. When Claude cites you, it is because your content met a higher evidentiary bar.
The operating implication: earned media placement in a single publication is not enough. Brands present on four or more third-party platforms are 2.8 times more likely to be cited in ChatGPT responses than brands on a single platform. Distributed earned media generates up to 325% more AI citations versus publishing on the brand site alone.
The game is not "get one placement." The game is build a citation network across independent sources.
What This Means for Your Content Investment
Most B2B companies spend the majority of their content budget producing owned content: blog posts, landing pages, whitepapers behind lead gates. The citation data says this strategy is structurally invisible to AI engines.
Here is the math. 42% of B2B decision-makers now use LLMs in their initial buying process. AI search visitors convert at 14.2% versus 2.8% from Google organic. ChatGPT referral traffic converts at approximately 16%, Claude at up to 16.8%.
The conversion rates from AI referrals are five to six times higher than traditional search. And 84% of the citations driving those referrals come from sources you neither own nor pay for. If your content strategy is 90% owned and 10% earned, you are optimizing for the channel that produces 5% of AI citations while ignoring the one that produces 84%. The measured unit is source composition in the cited sample. Boundary to preserve: this does not establish a universal source-selection mechanism, citation causation, a guaranteed future citation, recommendation lift, pipeline, revenue, or any other business outcome.
This is not a branding argument. It is an allocation error with measurable revenue impact.
The move: audit your content spend. Calculate what percentage goes to owned content versus earned media placement. Then look at your share of citation in AI responses for your core queries. If the ratio is inverted, you are funding the wrong channel.
How Machine Relations Changes the Equation
The traditional approach to earned media was relationship-driven and unmeasurable. You got a placement, you hoped it worked, you moved on. That model cannot compound.
Machine Relations treats earned media as an engineered system, not a networking exercise. The inputs are specific: which publications AI engines actually cite for your category queries, what entity chains connect your brand to the concepts buyers search, and what extractable claims make your content citable by machines.
The output is measurable: share of citation across specific queries, tracked over time. Not impressions. Not reach. Citation rate: how often an AI engine names your brand as the source when someone asks about your category.
The earned media that drives AI citations is not random press coverage. It is strategic placement in the publications and formats that each AI engine preferentially cites. It is entity chain architecture that connects your brand to the right concepts across multiple independent sources. It is structured content that machines can extract and verify.
This is the structural advantage that owned content alone cannot replicate. You can publish a thousand blog posts on your own domain. If none of them carry the third-party corroboration signal that AI engines weight most heavily, those posts remain invisible in the channel where 42% of your buyers now start their journey.
FAQ
What percentage of AI citations come from earned media versus owned content?
It depends which category the study counts, and the honest answer gives both. Muck Rack's analysis of more than 25 million AI-generated links puts sources brands neither own nor pay for at 84% of all AI citations, a figure that has held between 82% and 89% across three editions from July 2025 through May 2026. That bucket includes Wikipedia, Reddit, academic and government sources alongside journalism; journalism alone is 27%. Meltwater's narrower Earned/News category — press coverage and industry publications only, across eight engines — read 37.6% in May 2026, down from 39.5% in April. Brand-owned websites generate single-digit citation rates on every measure. The measured unit is source composition inside each study's observed citation sample. Boundary to preserve: Muck Rack Generative Pulse does not establish that earned media causes citation, that any one placement will be cited, or that earned media is a universal engine-selection mechanism.
Which AI search engine cites earned media the most?
ChatGPT cites sources in 96% of its responses, the highest rate among major AI engines. It averages 5 citations per response with Wikipedia as its top domain. Claude cites less frequently (55% of responses) but averages 13 citations when it does, favoring research-grade sources like PubMed Central.
Does distributing content across multiple publications increase AI citations?
Yes. Brands present on four or more third-party platforms are 2.8 times more likely to be cited in ChatGPT responses. Distributed earned media generates up to 325% more AI citations compared to brand-site-only publishing.
How do AI search referral conversion rates compare to Google organic?
AI search visitors convert at 14.2% versus 2.8% from Google organic. ChatGPT referral traffic specifically converts at approximately 16%, and Claude referrals at up to 16.8%. The traffic volume is smaller, but the conversion quality is five to six times higher.