Earned Media vs. Owned Content: AI Citation Rates Compared

Distributing content to third-party news publishers increases AI citation rates by 325%. That figure comes from a December 2025 study by Stacker and Scrunch, which analyzed 944 prompt–platform combinations across five AI platforms and found that the same article, when distributed across third-party news sites, raised citation rates from 8% to 34% — a 4.4x lift. Brand-only content achieved 7.6% citation rates. Earned distribution achieved 34%.

This is not a margin. It is a structural difference in how AI systems decide what to cite.

Earned sources receive more citations than owned content across all AI platforms

The Stacker/Scrunch study compared eight articles across different industries in two conditions: hosted only on the brand's domain, and distributed across hundreds of third-party news sites. Results were consistent across all eight topics, all five platforms, and all prompt types tested.

The 19.2% syndicated-only figure is the sharpest finding. In nearly 1 in 5 answers, AI systems cited the third-party version and did not cite the original brand piece at all. The publisher, not the brand, was treated as the credible attribution point for the same information.

A separate analysis by xFunnel.ai examined 40,000 AI responses containing 250,000 citations across ChatGPT, Google Gemini, and Perplexity, and found that earned content represents the largest percentage of citations by category across all three platforms — larger than owned brand content, competitor domains, or user-generated content (xFunnel.ai, 2025). The pattern held across all stages of the buyer journey, with third-party editorial content most concentrated at the problem exploration and solution education phases.

AI citations have almost no overlap with organic search rankings

The earned-versus-owned citation gap sits alongside a second structural finding: AI citations are nearly independent of organic search position.

Moz analyzed nearly 40,000 search queries in February 2026 and found that 88% of Google AI Mode citations are not in the organic SERP for the same query (Moz, 2026). Only 12% of AI Mode citations match the exact URLs appearing in Google's top 10. Even at the domain level, only 1 in 5 citations come from websites that appear in the top 10 at all.

That gap is not a flaw in the system. Moz attributes it to AI Mode's "fan-out" methodology, where a single user query triggers multiple related sub-queries simultaneously, and citations are aggregated from across all of them. A brand page ranking number 1 for a target keyword is one input into one sub-query in that aggregation.

The same decoupling appears in ChatGPT. Ahrefs analyzed citation overlap between ChatGPT's fan-out queries and Google's top 10 results and found only 6.82% URL overlap — fewer than 7 in 100 URLs that ChatGPT cites appear in Google's top 10 for the same queries (Ahrefs, September 2025). Profound's analysis of over 650 ChatGPT queries against 200 Google SERP pulls found 8–12% URL overlap depending on query type, with product queries showing a correlation of r ≈ –0.98 between ChatGPT citation frequency and Google rank — meaning the URLs ChatGPT favored for commercial queries were essentially the opposite of those Google ranked (Profound, 2025).

Optimizing exclusively for search rankings does not build AI citation authority.

What AI engines actually require to cite a brand

BrightEdge analyzed tens of thousands of prompts across ChatGPT, Perplexity, and Google's AI systems using its AI Catalyst tool and found that ChatGPT includes brand mentions in 99.3% of eCommerce responses — but only for brands it has already resolved (BrightEdge, October 2025). Resolution, in this context, means the AI system has encountered the brand across enough credible, independent sources to construct a stable entity association.

A brand that exists only on its own domain, without third-party coverage from publishers AI engines recognize as credible, is not a resolved entity. It is absent from the citation layer regardless of its domain authority or SERP position.

The BrightEdge data also shows that 48–77% of citations across all AI engines come from what the study calls "specialized sites" — industry publications, niche experts, and topic-specific resources outside the major platforms. This long tail of earned placement is where the majority of AI citations originate, not from brand-owned content.

Data density amplifies citation rates in both owned and earned content

The preference for earned sources operates independently from content structure, but content structure still determines whether a given piece — earned or owned — gets extracted when it is encountered.

The Princeton/Georgia Tech GEO paper (Aggarwal et al., SIGKDD 2024) tested 10,000 queries and found that adding statistics, quotations, and source citations each produced 30–40% improvement in AI visibility (Aggarwal et al., 2024). Keyword stuffing decreased visibility by 10%. The mechanism is direct: AI engines treat statistical claims and attributed data as reliability signals, affecting citation probability independent of the source domain.

Earned media already carries the third-party authority signal. Adding data density to earned content compounds the effect. The combination of trusted source, original statistics, and clear structure produces the highest citation rates across platforms.

Why AI engines systematically prefer earned sources

AI engines synthesize answers across multiple sources and cite the subset they assess as credible. That credibility assessment draws on signals that are structurally more available from earned third-party coverage than from brand-owned content.

A brand publishing on its own domain is self-asserting. A news outlet or industry publication covering that brand provides external corroboration. AI training data and retrieval systems weight corroboration over self-assertion — independent sources are harder to manufacture at scale and carry a different epistemic status in the training signal.

The Stacker/Scrunch study quantifies this at the measurement level. The 19.2% of citations going to syndicated-only sources represents AI engines actively selecting the publisher over the brand as the attribution point for the same information. The brand created the content. The publisher earned the citation.

This is the core inversion relative to traditional SEO. Google's algorithm rewarded the canonical source — the original, brand-owned piece. AI citation behavior rewards the most credibly corroborated source, which is typically the earned placement rather than the origin page.

Summary: the citation rate comparison in numbers

The data across studies is consistent: AI engines do not treat earned and owned content as equivalent sources. The gap is measurable, platform-independent, and tied to how AI systems construct credibility signals — not to any arbitrary platform preference.

Machine Relations — the discipline of earning AI citations by making a brand legible, retrievable, and credible across AI-driven discovery — treats earned media as the foundation layer precisely because the citation data demands it. More on the full framework at machinerelations.ai.