Press Releases Are Nearly Invisible in AI Search: What 4 Million Citations Reveal

For years, press release distribution services have pitched themselves as essential tools for search visibility. Now that AI search engines are reshaping how people discover information, these same services are rebranding their offerings with promises of "AI visibility." ACCESS Newswire offers an AI Visibility Checklist. eReleases published a guide positioning press releases as tools for AI search visibility. Business Wire has written about optimizing releases for answer engine discovery.

The pitch is seductive: write a press release, distribute it far and wide, and watch your brand appear in ChatGPT responses and Google AI Overviews. The reality, according to new data from a landmark citation study, tells a starkly different story.

The Study That Dismantles the Pitch

BuzzStream, in collaboration with Citation Labs, analyzed over 4 million AI citations across four major AI platforms: ChatGPT, Google AI Mode, Google AI Overviews, and Google Gemini. The research team ran 3,600 prompts across 10 industries and collected citation data for one full week. They used XOFU, a citation monitoring tool from Citation Labs, to track exactly where AI platforms pull their sources.

The methodology was rigorous. To identify syndicated content, BuzzStream cross-referenced author names against publications using its ListIQ tool and manually confirmed cases where the author name didn't match the publication. They acknowledged the method's limitations — some sites repost press releases without labeling them as such — but the scale of the dataset provides meaningful statistical power.

The findings are devastating for anyone banking on press releases for AI visibility.

The Numbers: Press Releases Are Statistically Invisible

News publications accounted for 14% of all AI citations in the dataset. That's not nothing — established journalism still matters to AI engines. But within that news category, syndicated and distributed content performs abysmally.

Press releases published through syndication channels like Yahoo and MSN accounted for 0.32% of news citations and 0.04% of the entire dataset. That is not a rounding error. That is functional invisibility.

Direct citations from newswire services like PRNewswire made up 0.21% of the full dataset. They appeared most often in exploratory and informational prompts, but even there they only reached 0.37% — still a rounding error in a dataset of millions.

Syndicated news content overall, including articles republished through MSN and Yahoo networks, accounted for 6.2% of news citations and 0.9% of the total dataset. Better than raw press releases, but still a tiny fraction of what original journalism captures.

To put this in perspective: for every 10,000 citations AI engines generate, approximately 4 will point to a press release distributed through syndication. The mathematical reality is that press releases, as a channel for AI visibility, are effectively dead.

Why AI Engines Ignore Press Releases

The data raises an obvious question: why do AI search engines systematically exclude press releases? Several factors likely contribute.

Training data quality filters. Large language models are trained on vast corpora of web content, but training pipelines typically include quality filtering mechanisms. Press releases, with their formulaic structure and promotional language, are relatively easy for content quality classifiers to identify and downweight. The same characteristics that make press releases feel inauthentic to human readers — excessive adjectives, lack of critical analysis, one-sided sourcing — make them easy for training pipelines to deprioritize.

Retrieval-time source selection. When ChatGPT or Google AI Overviews generate a response, they don't simply match keywords. They select sources based on authority, relevance, and informational value. A press release that announces a product launch with marketing language offers less informational value than a journalist's analysis of the same product. AI retrieval systems, which increasingly use relevance and authority signals similar to traditional search ranking, naturally favor the analysis over the announcement.

Economic incentives of syndication networks. Syndication networks like Yahoo News and MSN aggregate content from thousands of sources. AI engines appear to recognize the original source rather than the syndicated copy, similar to how Google's traditional search historically deduplicated syndicated content. A press release that appears on 50 syndication partner sites may count as a single low-value source rather than 50 signals of importance.

Structural format mismatch. AI engines synthesize information from multiple sources to construct answers. Press releases, by design, present a single perspective in a promotional frame. They are poor raw material for balanced synthesis. Journalistic articles, which typically include multiple viewpoints, expert quotes, and contextual analysis, are far better suited for AI answer construction.

What Actually Earns AI Citations

If press releases don't work, what does? The BuzzStream data, combined with other recent citation studies, points to clear patterns.

Original journalism from established publications dominates. The 14% of citations going to news publications is concentrated in original reporting from recognized outlets. Publications with editorial independence, fact-checking processes, and journalistic standards are preferred sources. The lesson is straightforward: if you want AI visibility, earn coverage from journalists who write original stories — not distribution services that replicate the same text across hundreds of sites.

Research and data-driven content performs exceptionally well. Studies, surveys, original datasets, and analytical reports frequently appear as AI citation sources. AI engines value content that presents novel information unavailable elsewhere. If your organization produces original research — industry surveys, market analyses, technical benchmarks — that content is far more likely to be cited than a press release announcing the research exists.

Expert analysis and opinion content earns citations. Thought leadership pieces from recognized experts, especially those published on established platforms, generate citations. AI engines appear to value named authors with demonstrated expertise over institutional bylines.

Comprehensive evergreen content. Detailed guides, reference pages, and comprehensive resources continue to perform well in AI citations, particularly for informational and exploratory queries. Content that thoroughly addresses a topic — with depth that a press release structurally cannot provide — is the kind of resource AI engines synthesize into answers.

The PR Industry's AI Visibility Illusion

The press release distribution industry has been quick to capitalize on AI visibility anxiety. Their marketing implies a causal relationship: distribute through our network, appear in AI answers. The BuzzStream data shows this relationship is essentially nonexistent.

This matters because marketing budgets are finite. Every euro spent on press release distribution services promising AI visibility is a euro not spent on strategies that actually work — original research, journalist relationships, expert content, and comprehensive resources.

The PR professionals who understand this shift are already pivoting. They're investing in media relations that produce original journalism rather than distribution that produces replication. They're creating data and research that journalists want to cover and AI engines want to cite. They're building author authority through consistent, expert content rather than relying on the reach of syndication networks.

Restructuring for AI Visibility

Organizations serious about appearing in AI-generated answers should treat the BuzzStream data as a strategic directive. The hierarchy of AI citation sources is becoming clear, and press releases sit near the bottom.

The path forward requires abandoning the distribution-first mindset. Instead of asking "how many sites will republish our announcement," the question becomes "what original information can we produce that AI engines will synthesize into answers."

This means investing in original research, cultivating relationships with journalists at publications that AI engines cite frequently, publishing expert analysis under named author bylines, and creating comprehensive resources that thoroughly address topics your audience cares about.

The data is unambiguous. In the emerging architecture of AI search, visibility is earned through originality and depth — not through the brute force of syndication. Press releases aren't completely useless; they still serve regulatory disclosure requirements and can inform journalists directly. But as a tool for AI search visibility, they are, according to 4 million data points, nearly invisible.