AI Visibility

Do Press Releases Help With AI Visibility? What 25 Million AI Citations Reveal

Earned media drives 84% of AI citations. Press releases account for less than 2%. Here's what 25 million citations across ChatGPT, Claude, and Gemini reveal about what actually builds AI visibility in 2026.

Jaxon ParrottMar 1, 2026

Do Press Releases Help With AI Visibility? What 25 Million AI Citations Reveal

Earned media drives 84% of all AI citations. Press releases represent less than 2%. If you paid $750 for wire distribution hoping ChatGPT would recommend your brand, the data says you chose the wrong mechanism.

This is not speculation. Muck Rack's Generative Pulse research — now in its third edition — analyzed more than 25 million links cited by ChatGPT, Claude, and Gemini across 17 industries. The pattern is consistent across every edition since July 2025: AI engines cite editorial journalism and earned media at rates that press releases cannot match, regardless of distribution volume.

The answer to "do press releases help with AI visibility?" is: marginally for trend-based queries, almost never for buyer-intent queries. Here is what the research shows, where the gap comes from, and what actually builds AI citation authority.

How Do AI Engines Decide What to Cite?

AI citation selection is not a content quality contest. It is a source reputation hierarchy.

A peer-reviewed study published at EMNLP 2025 — one of the top natural language processing conferences — found that for large language models, media source reputation matters more than content quality when generating citations. Researchers from Renmin University of China, the National University of Singapore, and the Chinese Academy of Sciences analyzed how LLMs select sources and confirmed that the name of the outlet — its established authority signal — is the primary driver of citation selection, not the content itself.

Separately, academic researchers analyzed over 366,000 citations from more than 24,000 conversations across OpenAI, Perplexity, and Google AI search systems (arXiv, 2025). Their finding: AI news citations concentrate heavily among a small number of outlets. You are competing against a narrow trust tier, not every indexed webpage.

This hierarchy was built during model training. Generative AI models absorbed internet-scale text data weighted toward journalism, encoding the sourcing norms and authority signals of editorial content before they ever processed a wire-distributed press release.

What Does the Research Show About Press Release AI Citations?

Press release citations grew in 2025 but hit a visible ceiling. The May 2026 Generative Pulse data makes the limits clear.

Muck Rack's longitudinal analysis — the only large-scale study tracking AI citation sources across multiple editions — shows:

Earned media accounts for 84% of all links cited by AI engines (consistent at 82–89% across three editions from July 2025 to May 2026)
Journalism alone makes up 27% of cited sources (stable at 25–27% across all editions)
Press releases appear at 3.5x higher rates in industry trend responses versus best-of queries — but remain marginal overall
Paid and advertorial content accounts for just 0.3% of AI citations
Journalistic sources represent approximately 49% of citations for queries that imply recency (breaking news, latest developments, timely topics)

Between July and December 2025, press release citations grew 5x — from 0.2% to roughly 1% of all AI citations. By May 2026, the growth curve had plateaued. Wire distribution gets indexed, but the content rarely reaches the citation tier that shapes AI-generated answers about brands, products, or services.

Why Does Earned Editorial Coverage Dominate AI Citations?

Three structural reasons explain the 84% earned media advantage, and none of them are fixable with better press release writing.

Training data bias. Generative AI models were trained on internet-scale text heavily weighted toward journalism. Reuters, the Financial Times, Forbes, Axios, and similar outlets are overrepresented in training corpora. The models internalized editorial authority signals before they processed a single syndicated press release.

Third-party validation signal. AI engines trained on journalism have learned to distinguish between content chosen by an independent editor and content a company self-published with wire distribution. A Forbes journalist covering your product independently carries an editorial validation signal. A PR Newswire release does not.

Source concentration. When the Nieman Journalism Lab analyzed which outlets ChatGPT and Gemini cite most, the pattern was definitive: Reuters, the Financial Times, Time, Forbes, and Axios consistently appeared in the top tier (Nieman Lab, July 2025). Axios alone appeared in the top three cited domains across 13 of 17 industries tracked on ChatGPT. Over 20,000 distinct journalism outlets were cited across the dataset — but citations concentrate among a small, identifiable tier.

These are structural dynamics, not algorithm quirks. You cannot overcome them by improving your press release format, adding schema markup, or choosing a different wire service.

How Does Wire Distribution Fail for AI Visibility?

The wire distribution mechanism is specifically mismatched with how AI engines build citation trust.

When you distribute via PR Newswire, Business Wire, or GlobeNewswire, the primary copy lands on the wire service's site and syndicates to hundreds or thousands of local newspaper sites, aggregators, and republishing partners. The EMNLP 2025 research finding applies directly: AI models cite based on outlet reputation, not content quality. A well-written release on a local news aggregator does not carry the authority signal of the same story covered independently by a Financial Times journalist.

The mechanism fails in three specific ways:

No third-party editorial validation. Wire distribution is brand-generated, brand-paid content. AI engines trained on journalism recognize this distinction.
Syndication reads as spam, not diversity. Thousands of sites republishing identical release text looks like content duplication to AI crawlers, not editorial consensus.
Outlet tier mismatch. Even professional wire distribution lands content in a citation tier far below a brief mention in Reuters, Forbes, or Axios.

The May 2026 Muck Rack data captures this precisely: despite growth in press release indexing, earned editorial coverage holds at 84%. The gap is structural.

Which Publications Do AI Engines Actually Cite Most?

Citation authority concentrates among a specific, identifiable tier of outlets — and the tier varies by AI platform.

The May 2026 Generative Pulse data reveals that each major AI platform has distinct citation preferences:

AI Platform	Citation Rate	Avg Citations per Response	Top Cited Domain	Journalism Reliance
ChatGPT	96% of responses include citations	5 per response	Wikipedia	Highest for recency queries
Gemini	82% of responses include citations	8 per response	Reddit	Balanced editorial/community
Claude	55% of responses include citations	13 per response	PubMed Central	Academic/research-weighted

This platform divergence matters for PR strategy. ChatGPT — the platform with the highest consumer adoption — cites journalism most aggressively for recency queries. Gemini leans toward community-validated content alongside editorial sources. Claude indexes more heavily toward academic and government sources.

The academic citation pattern research confirms the concentration effect: the top outlets capture a disproportionate share of AI news citations, while smaller publications — even legitimate editorial operations — rarely appear in AI-generated responses. The publications that get cited most by AI search engines in 2026 are a specific tier: tier-one business press, major tech outlets, and specific industry authorities.

Your PR strategy needs to target those outlets specifically, not "media coverage" generally.

Press Releases vs Earned Media vs Owned Content: AI Citation Comparison

The data from Muck Rack, the EMNLP 2025 study, and the arXiv citation analysis converge on a clear hierarchy. Here is how each content type performs for AI citation authority:

Content Type	Share of AI Citations	AI Trust Signal	Time to Citation Impact	Best Use Case
Earned editorial coverage (journalist-chosen stories in trusted outlets)	~84%	Highest — third-party editorial validation, outlet reputation	2–6 months of relationship building	Category authority, buyer-intent queries, brand recommendations
Journalism in tier-one outlets (Reuters, FT, Forbes, Axios)	~27% of all citations (49% for recency queries)	Very high — top of AI training hierarchy	Ongoing media relationship investment	Breaking news, trend queries, industry authority
Press releases (wire-distributed)	<2% of all citations	Low — brand-generated, no editorial validation	Immediate indexing, minimal citation impact	Newsworthy announcements that may trigger earned coverage
Paid/advertorial content	0.3%	Negligible	Immediate placement	Not effective for AI visibility
Brand-owned content (company blog, website)	Small minority	Low-moderate — no third-party validation	Ongoing SEO investment	Owned-query defense, technical documentation
Academic/research content	Varies by platform (high on Claude)	High — institutional credibility	Publication cycle dependent	Research-backed authority claims

The hierarchy is clear: editorial validation from trusted outlets drives AI citation authority. Distribution volume without editorial selection does not.

What Actually Builds AI Citation Authority?

The research converges on a specific model that separates effective strategies from wasted budget.

AI citation authority = editorial placements in the specific publications AI engines trust

This requires:

Media relationships with tier-one outlets. Not a media list — actual journalist relationships that produce original editorial coverage in Reuters, Forbes, Axios, the Financial Times, and industry-specific authorities.
Stories worth covering independently. AI engines cite content that journalists chose to write about, not content companies paid to distribute. Your story needs editorial news value.
Consistent presence, not single placements. AI models build brand knowledge from repeated editorial mentions across trusted outlets over time. A single Forbes article helps; sustained coverage across multiple tier-one outlets compounds.
Extractable, structured claims. Research from Omniscient Digital's analysis of 23,000+ AI citations found that 44.2% of all LLM citations come from the first 30% of page content (Omniscient Digital, 2026). Content structured around quotable data points, clear definitions, and extractable tables gets cited more than higher-authority pages without those properties. Pages built with answer-first paragraphs, descriptive headings, and comparison tables are cited 2–4x more frequently across ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews (Kime.ai, 2026).
Original data and research. Original research and data-rich reports are cited at 3–10x the rate of standard content (Leapd, 2026). Proprietary data creates a citation moat. Princeton and KDD research on Generative Engine Optimization confirmed that adding statistics to content improves AI visibility by up to 41%, while adding quotations lifts citation rates by 32% (OmniBound, 2026).

What does not work for AI citation authority:

Press wire distribution for routine announcements
Guest posts on brand-selected sites (no independent editorial validation)
Thought leadership on LinkedIn or newsletters (limited AI indexing)
Content designed for AI optimization without editorial validation behind it

This is the model earned media dominates AI search results research confirms: the right editorial relationships in the right outlets, not more distribution volume.

How to Evaluate Your 2026 PR Budget for AI Visibility

The citation data suggests a clear framework for allocating PR resources when AI visibility is a goal:

High ROI for AI visibility:

Direct editorial placements in the tier-one outlets ChatGPT, Gemini, and Claude cite most (Reuters, Financial Times, Forbes, Axios, and industry-specific authorities)
Media relationships that produce original journalist-driven coverage
Original research and proprietary data that creates citation-eligible assets
Consistent earned media presence across the outlets in your ICP's AI search results

Low ROI for AI visibility:

Press wire distribution for routine announcements (product launches, standard hiring news, sub-$50M funding rounds)
Syndication-heavy distribution strategies that prioritize volume over outlet quality
Paid/advertorial placements (0.3% of AI citations)
Content optimized for AI extraction without underlying editorial authority

Conditional ROI:

Press releases tied to genuinely newsworthy events — major funding, significant partnerships, original data — that may trigger independent editorial coverage
Wire distribution to tier-one services (PR Newswire, Business Wire) as a complement to editorial outreach, where the wire's own site provides some indexing value

Press releases are not useless. They serve corporate communications, investor relations, and regulatory disclosure. But for AI citation authority specifically, they represent less than 2% of the system. The path to AI citations runs through earned media strategy, not distribution technology.

The Machine Relations Framework

The reason this matters goes beyond visibility metrics. AI search is now embedded in B2B buying workflows. When a buyer opens ChatGPT or Perplexity to research solutions, they see a curated set of brands — the ones editorial institutions validated and AI models learned to trust.

If you are not in that set, you are invisible at the exact moment a prospect is building their shortlist.

The data across three editions of Muck Rack's Generative Pulse, the EMNLP 2025 peer-reviewed research, and multiple academic citation analyses tells a consistent story: AI citation authority is earned, not distributed. It comes from being chosen by journalists at outlets AI engines trust, not from wire-service syndication volume.

Machine Relations is the systematic practice of building that earned media presence — in the right outlets, producing the citation signals AI engines use to recommend brands to buyers who are actively searching.

Companies winning AI visibility in 2026 are not the ones with better wire distribution. They are the ones that built genuine editorial relationships in the publications AI was trained to trust.

FAQ

Do press releases help with AI visibility?

Press releases contribute marginally to AI visibility. Muck Rack's May 2026 Generative Pulse analysis of 25 million citations found press releases appear at 3.5x higher rates in industry trend queries versus best-of queries — but earned media accounts for 84% of all AI citations while press releases represent less than 2%. Wire distribution alone does not build the editorial authority AI engines require for citation selection.

What percentage of AI citations come from earned media?

Earned media accounts for 84% of all links cited by AI engines, according to Muck Rack's May 2026 Generative Pulse report analyzing citations across ChatGPT, Claude, and Gemini. This figure has been consistent at 82–89% across three editions of the study from July 2025 to May 2026. Journalism alone makes up 27% of cited sources.

Which AI platform cites the most sources?

Claude cites the most sources per response at 13 citations per answer, though it includes citations in only 55% of responses. ChatGPT includes citations in 96% of responses at 5 per answer. Gemini cites 8 sources per response in 82% of answers. Each platform favors different source types: ChatGPT leans toward Wikipedia and journalism, Gemini toward Reddit and editorial content, and Claude toward PubMed Central and academic sources.

Which publications do ChatGPT and Gemini cite most?

Reuters, the Financial Times, Time, Forbes, and Axios consistently appear in the top-cited tiers for both ChatGPT and Gemini, according to Nieman Lab reporting on Muck Rack data. Axios appears in the top three cited domains across 13 of 17 industries tracked on ChatGPT. Over 20,000 distinct journalism outlets appear in the dataset, but citations concentrate heavily among tier-one outlets.

Why do AI engines favor editorial coverage over press releases?

Peer-reviewed research published at EMNLP 2025 found LLMs prioritize media source reputation over content quality when generating citations. AI models trained on internet-scale text weighted toward journalism internalized editorial authority signals. Press releases are brand-generated content without independent editorial validation — the third-party credibility signal AI engines were trained to recognize and cite.

How long does it take to build AI citation authority?

AI citation authority builds through consistent earned media placements over months. Unlike wire releases that index quickly but generate minimal lasting citation signals, editorial coverage in tier-one publications creates cumulative presence AI engines incorporate into brand knowledge. Companies with sustained earned media presence across Reuters, Forbes, Axios, and industry authorities see compounding citation effects over 3–6 months.

Sources

Muck Rack. "Earned Media Still Drives 84% of AI Citations. Here's What That Means for PR." Generative Pulse, May 2026. https://muckrack.com/blog/what-is-ai-reading-may-2026
Muck Rack. "Generative Pulse: Earned Media Consistently Drives AI Citations, Holding at 84%." GlobeNewswire, May 7, 2026. https://www.globenewswire.com/news-release/2026/05/07/3290268/0/en/Generative-Pulse-Earned-Media-Consistently-Drives-AI-Citations-Holding-at-84.html
Dai, S., Cao, Z., et al. "Media Source Matters More Than Content: Unveiling Political Bias in LLM-Generated Citations." Proceedings of EMNLP 2025, Association for Computational Linguistics. https://aclanthology.org/2025.emnlp-main.872.pdf
Yang, K. "News Source Citing Patterns in AI Search Systems." arXiv, 2025. https://arxiv.org/html/2507.05301v1
Minici, M., et al. "Auditing LLM Editorial Bias in News Media Exposure." arXiv, October 2025. https://arxiv.org/html/2510.27489v1
Deck, Andrew. "Generative AI Models Love to Cite Reuters and Axios, Study Finds." Nieman Journalism Lab, Harvard University, July 2025. https://www.niemanlab.org/2025/07/generative-ai-models-love-to-cite-reuters-and-axios-study-finds/
Muck Rack. "Earned Media Still Drives Generative AI Citations as Press Release Visibility Grows." GlobeNewswire, December 2, 2025. https://www.globenewswire.com/news-release/2025/12/02/3198248/0/en/Earned-Media-Still-Drives-Generative-AI-Citations-as-Press-Release-Visibility-Grows.html
Nifong, Casey. "How Generative Engines Define and Rank Trustworthy Content." Search Engine Land, September 5, 2025. https://searchengineland.com/how-generative-engines-define-rank-trustworthy-content-461575
Muck Rack. "What Is AI Reading? Generative Pulse Report." https://generativepulse.ai/report/
Omniscient Digital. "Which Content Types LLMs Cite Most: 23,000+ AI Citations Analyzed." 2026. https://beomniscient.com/blog/content-types-cited-in-llms/
Kime.ai. "How to Structure Content for LLM Extraction: A GEO Guide for 2026." 2026. https://kime.ai/blog/how-to-structure-content-for-llm-extraction-geo-guide-2026
Leapd. "How ChatGPT, Google AI Overviews, and Perplexity Source Information in 2026." 2026. https://www.leapd.ai/blog/ai-visibility/how-chatgpt-google-ai-overviews-and-perplexity-source-information-in-2026
OmniBound. "Generative Engine Optimization Statistics (2026): 60+ Data Points on AI Citations, Brand Visibility, and Content Performance." 2026. https://www.omnibound.ai/blog/generative-engine-optimization-statistics