Should You Invest in Earned Media for AI Citations? Yes — In This Order

Yes: fund earned media and third-party distribution before you fund more on-site content for AI visibility, but do not defund owned content to do it. That is the operating decision behind six separate 2025-2026 studies, five of which measured AI engines citing third-party and earned sources against brand-owned pages for discovery and category questions and one of which (GEO-16) audited on-page quality and recommends pairing it with third-party coverage: Muck Rack's Generative Pulse analysis of more than one million cited links, Stacker and Scrunch's first-look distribution study, Ahrefs' correlation study of 75,000 brands, Chen et al.'s large-scale GEO study (arXiv:2509.08919), Kumar et al.'s GEO-16 framework (arXiv:2509.10762), and Fullintel and UConn's peer-reviewed sample. The full evidence base, with each study's method and limits, lives at Machine Relations' earned-media citation research. This page is the decision that follows from it.

The Machine Relations Index measures the same question by source role, and it settles which kind of earned media to buy. On buyer questions in its October 19, 2026 release (Index methodology v2.1; 20,868 answer runs on six buyer question shapes across 19 categories, May 10 to October 19, 2026), editorial publications took 30.0 percent of the citations that landed on a role-classified domain, company websites 25.8 percent, and wire and press-release distribution 0.8 percent. Editorial publications were cited 37 times as often as wire domains. The earned line that matches where engines look is coverage on publications, not a release on a wire; and company websites still take a quarter of classified citations, which is why owned content stays funded. Those shares rest on the 28.6 percent of buyer-question citations that land on a classified domain, and they describe what engines cited, not what any placement produces.

The full source-role table, by question shape and for the whole release, with what it does not establish: earned versus owned AI citation rates, ranked.

Method note. Every Index figure on this page can be recomputed from the public release artifact at machinerelations.ai/data/machine-relations-index.json (release mri_score_v2.1+2026-10-19+da0ee4378219): sum each domain's runs_cited over the published strata with the best-of, how-to-choose, is-it-worth-it, problem-first, top-list and X-versus-Y question shapes, group by source_role, and divide by the total excluding Other observed source.

The sequencing matters more than the headline. A brand that moves its whole content budget into press placements and lets its owned pages go stale loses the entity clarity and factual extraction that AI systems still need from the brand's own site. A brand that keeps investing only in owned content without building credible third-party presence is optimizing one part of a citation surface where, per McKinsey's AI Discovery Survey, brand-owned websites account for only 5-10% of the sources AI search references. Both failure modes are common. Neither is what the research supports.

What to do, in order

• Keep owned content live and extractable. It still supplies the factual base AI systems compare the earned-media graph against — do not cut it to fund placements.
• Prioritize brand mentions over backlinks. Ahrefs found brand web mentions correlate three times more strongly with AI Overview visibility than backlinks (0.664 vs. 0.218). Editorial coverage and thought-leadership placements earn mentions; pure link-building does not.
• Distribute earned stories widely, not once. In Stacker and Scrunch's eight-story first-look pilot, about 34% of answers cited some version of the stories when they ran across many third-party outlets, against about 8% citing only the brand's original; a follow-up study found a 239% median lift across a broader sample. One placement is a weaker bet than the same story syndicated. Distribution here means the story running on publications and third-party sites, not a wire release: in the Index, wire and press-release domains take 0.8 percent of classified buyer-question citations.
• Measure citation, mention, and revenue as separate outcomes. None of these studies show that a mention causes a citation, or a citation causes pipeline. Track each with its own number, by brand, host, URL, query, and engine.

The framework that keeps those four actions coordinated instead of competing for budget is Machine Relations, which I coined in 2024 as the discipline of making a brand legible, retrievable, and credible inside AI-driven discovery, with earned authority as one of five layers alongside entity clarity, citation architecture, distribution, and measurement.

The evidence behind that decision, in brief

Six separate studies point the same direction without proving identical claims. Muck Rack's report found its broad earned-media category made up most citations in its observed sample. Stacker and Scrunch's test, run by Stacker, which sells the distribution it measured, found about 34% of answers citing some version of eight stories syndicated across third-party outlets against about 8% citing only the brand's original. Ahrefs found brand web mentions correlate far more strongly with AI Overview visibility than backlinks. Chen et al. (University of Toronto) found AI search shows a "systematic and overwhelming bias towards Earned media... over Brand-owned and Social content." Kumar et al.'s GEO-16 audit found on-page quality strongly associated with citation and recommended pairing it with off-site coverage, while noting it did not test publication venue. Fullintel and UConn's sample found journalistic and unpaid sources made up large shares of cited links.

None of these studies establish a universal earned-media share, a provider-selection mechanism, a single-placement guarantee, or a direct revenue result — each measured source composition, correlation, or citation-rate change inside its own sample. That distinction, study by study, with full quotes and methodology notes, is what Machine Relations' research page exists to hold. Read it before you cite a specific percentage to a client or a board. The practitioner side confirms the same direction: WorldCom Group, a consortium of 160 PR agencies, and Edelman's U.S. media head in Campaign Asia are both describing earned media as central to AI visibility work, and PR Daily's 2026 predictions treat AI citation as a primary comms metric. Search Engine Land's GEO guide tells the same story from the SEO side: earned media is now a direct GEO lever, not a separate discipline.

Two further data points shape timing, not direction. Forrester's 2026 Buyer Insights data found 94% of B2B buyers now use AI in their purchasing process, so this is a current discovery concern, not a future one. And Zhang et al. (arXiv:2512.09483) found 37% of AI-cited domains never appear in traditional search results at all — Moz's 2026 analysis of 40,000 queries found most Google AI Mode citations do not appear in the organic top 10, so SEO wins cannot stand in for this work.

What this means for how you spend

Your own website is a minority citation source for AI

McKinsey's data indicates brand-owned websites may be a minority source for some AI-search reference behavior. That limits what on-site optimization alone can accomplish. A brand that has invested entirely in SEO and owned content has built a strategy around only one part of the citation surface — directional source-mix evidence, not a universal 90-95% rule.

The Index shows the other side. Company websites take 25.8 percent of classified buyer-question citations and come within three points of editorial publications on four of the six buyer question shapes. Every company's site is classed as vendor-owned, so for any single brand most of that share is other companies' sites: competitors, suppliers and platforms. Your own site is a small slice of the answer, not an empty one, and it is the page engines check your facts against.

Backlinks are not the same budget line as brand mentions

Backlinks are acquired through content and technical relationships. Brand mentions are acquired through editorial coverage, media placements, and thought leadership — the domain of PR, not link building. If your AI-visibility budget is still mostly an SEO link-building line, it is funding the weaker correlation.

Distribution is the lever, not the single placement

In the Stacker and Scrunch data, the republished copies were what got cited; the tested treatment was the same story running across many third-party domains, not one bespoke placement. Budget for syndication and multi-outlet pickup on publications, not a single press release: the six buyer question shapes in the Index cited prnewswire.com, the most-cited wire domain, in 95 of 20,868 runs; Christian Lehman maps the implementation layer for converting press placements into citation-ready sources.

Structure helps extraction; it does not substitute for earned authority

Pages with strong metadata and freshness signals, semantic HTML and structured data were cited more often in the GEO-16 study (Kumar et al., 2025), an observational audit of 1,100 URLs cited by Brave Summary, Google AI Overviews and Perplexity. That is a real, cheap lever — but it is a multiplier on earned authority, not a replacement for it.

Next Step

Start your visibility audit to see which publications engines cite in your category before you set the earned-media budget.

Frequently asked questions

Should I cut owned content spend to fund earned media?

No. McKinsey's data shows owned websites are a minority AI-citation source, not a zero one — brand pages still supply the factual base AI systems compare third-party coverage against. Fund both; shift new budget toward earned distribution rather than defunding what you have.

What share of AI citations come from earned media?

It depends on what a study counts. On buyer questions in the October 19, 2026 Machine Relations Index release, editorial publications took 30.0 percent of role-classified citations; the widely quoted 84 percent earned-media figure counts every non-owned, non-paid source, not journalism alone. For the exact figures each study reports, and what each one does and does not establish, see Machine Relations' earned-media citation research.

Does improving SEO rankings improve AI citation rates?

Not automatically. Zhang et al. (arXiv:2512.09483) found AI-cited domains absent from traditional search entirely, and Moz's analysis found most Google AI Mode citations do not appear in the organic top 10. Treat SEO and AI citation as separate budget lines with separate measurement.

Is a press release on a wire the same as earned media for AI citations?

No. On buyer questions in the October 19, 2026 Machine Relations Index release, wire and press-release distribution domains took 0.8 percent of role-classified citations and editorial publications took 30.0 percent. Treat a wire release as an announcement channel; fund coverage on the publications engines cite in your category.

Where to go next, by the question you are asking

This page answers the budget decision. The full evidence base is at Machine Relations' research page. Eight other pages answer the execution decisions that follow, and each one is the primary answer for its own question.

• How to Optimize Earned Media for GEO: 7 Tactics That Get AI Citations: the execution playbook, including whether original research belongs in the mix.
• How to Get Cited in AI Search: getting cited by a named engine, Gemini, Perplexity, ChatGPT or Claude.
• Citation Economy: How Earned Media Shapes AI Visibility: what citation rate means and how to read it.
• How Earned Media Now Dominates AI Search Results: whether earned media improves AI search results, and the best earned media strategies for AI search.
• Best AI PR Agencies 2026: Which Earned Media Agency Gets Your Brand Cited in AI Search: which earned media agency to hire for AI search.
• GEO Earned Media Strategy: how to split a GEO programme between owned and earned media.
• PR for AI Search: running earned media for answer engines, and what to ask an agency for.
• LLM SEO: Earned Media Beats On-Page Optimization: whether earned media or on-page work builds durable LLM visibility.