How Earned Media Drives AI Search Visibility: What the Machine Relations Index Shows

Earned media is the single largest driver of AI citations. When ChatGPT, Perplexity, or Gemini answers a buyer's question and names a brand, the citation almost always traces back to a third-party publication, not the brand's own website. That is well supported by outside research, and the Machine Relations Index shows which publications carry it.

What the Index shows

Measured on one denominator, Machine Relations Index release mri_score_v2.0+2026-09-20+47973f373a20, 16,039 monitored answer runs from 2026-05-10 to 2026-09-20 across six engines, with an evidence floor of 10 observed runs on 7 distinct dates:¹

Publication	Citation rate	Cited runs	Days cited	Engines	Measured standing
medium.com	5.04%	808	109 of 127	6	rank 1 of 1,230
forbes.com	4.15%	665	106 of 127	6	rank 2 of 1,230
prnewswire.com	1.13%	182	75 of 127	6	rank 8 of 1,230
techcrunch.com	1.02%	164	44 of 127	5	rank 9 of 1,230
techbullion.com	0.07%	12	11 of 127	3	below the floor for a rank

How to read the editorial-class rank. The editorial-class rank above is read off the Machine Relations Index's source_role field, which in the current release combines source-type evidence with AuthorityTech's placement catalog; 96 of the 1,231 class members are admitted through the catalog, and the Index is moving editorial classification to source-type evidence alone. The class currently includes six platform domains alongside edited newsrooms: medium.com (class rank 1), amazon.com (7), prnewswire.com (8), apple.com (16), wordpress.com (239) and blogspot.com (330). At the same time substack.com, beehiiv.com and dev.to — the same hosted open-publishing product as Medium — are filed as community platforms. Those six hold 1,358 of the editorial class's 13,525 cited runs, 10.04%, and they include the class's top slot. The class denominator also moves between releases (published figures carry 1,025 to 1,230) because it counts the domains classified in each release. What is unaffected: the citation rate, cited runs, days cited, engine breadth and confidence on this page are direct per-domain observations over a fixed run set, and are the primary measure; read the rate first.

Forbes is cited 3.65 times as often as the largest wire. On a rate, what ranks highest is whoever wins the most answers, not whoever publishes the most URLs.

How many publications get cited. The current release observes 22,320 distinct cited domains across 16,039 answer runs. The universe of publications that get cited at all is large; what is concentrated is recurrence. Most publications are cited too rarely to clear an evidence floor, which is a statement about measurement thresholds, not about exclusion by the engines. An absence claim needs engine breadth behind it, because a source not observed by one crawler may be cited plenty by the others.

What this means for prestige. Medium at rank 1 and Forbes at rank 2, both cited on more than 100 of 127 days across all six engines, show that neither prestige alone nor volume alone decides it. Distribution breadth and recurrence matter, and the third-party research below makes that case directly.

Why most publications are cited too rarely to rank

The concentration is real. AI engines develop citation preferences based on which publications appear most frequently and most reliably across query domains. A publication has to clear a recognition threshold before it becomes a citation candidate.

Three factors determine whether it clears that threshold:

1. Cross-domain corroboration. Research from the University of Toronto found that AI engines cite earned media 5x more frequently than brand-owned content, with 82-89% of all AI citations originating from third-party publications (University of Toronto, 2026). The mechanism: AI systems build citation confidence when multiple independent sources discuss the same brand in the same context. One Forbes mention does not build that signal. Forbes plus TechCrunch plus a regional tech publication plus a wire pickup: that cluster is what creates citation eligibility.

2. Distribution breadth. Downstream editorial amplification matters. When a story gets picked up by 40 regional outlets, each creating an independent indexed page about the same brand, AI engines see 40 corroborating sources instead of one. Medium's measured position — the most-cited editorial publication in the release, cited on 109 of 127 observed days across all six engines — reflects that dynamic. The wire does not: measured, it sits eighth of 1,230.

3. Query surface breadth. Publications cited across many query types carry more citation weight than those cited in a narrow vertical. A publication appearing in answers to enterprise software queries, fintech queries, and cybersecurity queries simultaneously builds a wider recognition surface than a high-prestige outlet with a single-category focus.

The PR strategy misalignment that kills earned citations

Muck Rack's May 2026 "What Is AI Reading?" report analyzed more than 25 million cited links across ChatGPT, Claude, and Gemini and found that earned media accounts for 84% of all AI citations, a figure that has held between 82% and 89% across every edition since July 2025 (Muck Rack, May 2026). The pattern is not tied to a particular model update or time period. It is structural.

Most PR campaigns optimize for prestige placement without accounting for the distribution mechanics that turn placements into earned citations.

A placement in a publication that gets amplified across many downstream outlets can generate more AI citation surface than a single placement with no amplification chain. The controlled evidence for that is the Stacker and Scrunch experiment below.

Stacker and Scrunch ran the largest controlled GEO experiment published to date: 87 earned media stories across 30 brands, queried across 2,600+ prompts on 8 AI platforms (Stacker/GlobeNewswire, March 2026). The median lift from earned media distribution was 239%. Ninety-seven percent of distributed stories earned at least one AI citation. The driver was distribution breadth, not the prestige tier of the placement.

5WPR's "Who AI Cites Now" report found that brands appearing on four or more third-party platforms are 2.8x more likely to be cited in ChatGPT responses than single-platform brands (5WPR, 2026). That is not a content quality variable. It is a distribution architecture variable.

How earned media creates AI citation eligibility

Machine Relations, which I coined in 2024, is the discipline of earning AI citations and recommendations for a brand by making that brand legible, retrievable, and credible inside AI-driven discovery. The distinction between Machine Relations and traditional PR is not about tactics. It is about the success condition.

Discipline	Optimizes for	Success condition	Scope
SEO	Ranking algorithms	Top 10 position on SERP	Technical + content
GEO	Generative AI engines	Cited in AI-generated answers	Content formatting + distribution
AEO	Answer boxes / featured snippets	Selected as the direct answer	Structured content
Digital PR	Human journalists/editors	Media placement	Outreach + storytelling
Machine Relations	AI-mediated discovery systems	Resolved and cited across AI engines	Full system: authority, entity, citation, distribution, measurement

The MR Stack positions earned authority at the foundation layer because AI engines cite third-party sources at 4-6x the rate of brand-owned content. You cannot build citation eligibility from your own blog. You have to earn it in the publication ecosystem, specifically in the publications AI engines have already decided to trust.

How to build your earned citation architecture

Citation concentration is workable. But it requires treating earned citations as a target metric rather than assuming they emerge naturally from quality PR.

Start here: run a baseline citation audit before any campaign. Not where you think your brand should appear in AI answers. Where it actually appears. Query the AI engines your buyers use with the questions they actually ask about your category. See whether your brand appears, whether competitors appear instead, and which publications are being cited in those answers.

That gap between current citation share and your competitive position tells you what the earned media architecture has to accomplish. Which publication clusters need your brand in them. Which distribution channels connect to the outlets AI engines already trust in your vertical.

AuthorityTech's visibility audit runs this analysis across the AI engines relevant to your category. The output shows which publications currently drive citation share in your space, which ones are within reach for earned media, and what the distribution pattern looks like for brands that have already cleared the citation eligibility threshold.

Publications and brands that never surface in AI answers are not usually there because they made the wrong bets. They are there because nobody mapped the citation architecture before they started spending.

---

Frequently asked questions about earned media and AI citations

What are earned citations in AI search?

Earned citations are mentions of a brand in AI-generated answers that trace back to third-party earned media placements rather than brand-owned content. When Perplexity or ChatGPT cites a brand in response to a buyer query, the source is almost always a third-party publication. The University of Toronto found that 82-89% of all AI citations originate from third-party sources (arxiv.org). Earned citations are the measurable output of earned media in AI search.

How does earned media drive AI citations?

Earned media drives AI citations by creating the distributed third-party corroboration signal that AI engines require before citing a brand. AI systems build citation confidence when multiple independent sources discuss the same brand in the same context. A single placement rarely triggers citation eligibility. A cluster of placements across wire pickups, trade publications, and indexed platforms creates the recognition surface AI engines need to include a brand in answers.

Why do some publications get far more AI citations than others?

Citation frequency reflects distribution breadth, content density, cross-domain relevance and recurrence rather than editorial prestige alone. Measured across 16,039 monitored answer runs on six engines from 2026-05-10 to 2026-09-20, Medium is cited in 5.04% of runs and Forbes in 4.15%, while PR Newswire is cited in 1.13%. AI systems develop citation preferences through patterns in training data and retrieval indices: publications appearing frequently across many query domains build stronger citation signals than high-prestige outlets with narrower distribution footprints.

What is Machine Relations and how does it connect to earned citations?

Machine Relations, coined by Jaxon Parrott, founder of AuthorityTech, in 2024, treats AI citation eligibility as a systematic architecture problem. The question it asks: has a brand built the earned authority, entity resolution, and distribution coverage that AI discovery systems require to cite it? Share of citation, the percentage of relevant AI responses that name your brand, is the primary metric. Earned citations from earned media placements are the foundation of that share.

Sources and method

The University of Toronto analysis, Muck Rack's "What Is AI Reading?" May 2026 report, the Stacker and Scrunch controlled experiment, and the 5WPR study are linked at their own sources; they are the evidence this post's thesis rests on. The engines publish what they crawl and how to control it: Google's common crawlers, OpenAI's bots, Perplexity's bots, Anthropic's crawler policy. A citation claim should meet that same standard of specificity. Ask any leaderboard for its denominator, its window, its engine breadth and its evidence floor. Ask an absence claim the same questions twice, because absence is the easiest artifact to mistake for a finding.

Updated 2026-09-23: figures reflect the current Machine Relations Index methodology; see the MRI methodology and update log.

Notes

Footnotes

1. Machine Relations Index, release mri_score_v2.0+2026-09-20+47973f373a20, window 2026-05-10 to 2026-09-20, 919 monitored prompts, 16,039 answer runs, 22,320 cited domains, six engines. Read live 2026-09-20 at https://machinerelations.ai/index and the linked domain profiles. ↩