Do Press Releases Get Cited by AI? A Study

Press releases have been a cornerstone of brand communication for decades, but in the age of AI search, their fate is more complicated than most marketers realize. Two studies sit in apparent contradiction: Loganix ran 100 queries across major AI platforms and found zero press release wire domains cited in any response. Meanwhile, Muck Rack's earned media research found that wire-distributed PR citations in AI answers grew five times over between July and December 2025. Both findings are accurate, and understanding why requires pulling apart what's actually happening when AI engines decide what to cite.

The short answer: it's not the press release that gets cited. It's the media coverage the press release generates. The PR is the trigger. The citation source is the journalist, the tech publication, the trade outlet that picked up the story. Once you internalize this distinction, both data points make perfect sense, and you can build a fundamentally better PR strategy around it.

Why Most Press Releases Don't Get Cited by AI

AI language models are trained to synthesize authoritative, informational content. Press releases, structurally, are promotional documents. They're written to announce something favorable to the issuing organization. That framing creates several problems for AI citation selection.

First, wire domains lack topical authority. PR Newswire, Business Wire, and GlobeNewswire are distribution platforms. They publish content from thousands of organizations across every conceivable industry. From an AI model's perspective, these domains have no deep expertise in any particular subject, they're content aggregators with low editorial selectivity. AI engines like ChatGPT and Gemini appear to weight domain topical authority significantly when selecting sources, and wire services simply don't accumulate it.

Second, press releases aren't primarily informational. The content of a typical press release, even a well-written one, is structured around announcement framing: "Company X today announced…" That structure signals promotional intent. AI models trained on large corpora of text learn to distinguish informational writing from marketing writing, and they deprioritize the latter as a citation source even when the underlying facts are accurate.

Third, the source is inherently non-independent. AI engines, particularly when answering factual or evaluative questions, seem to weight independent sources over self-reported company claims. A press release saying your product is "the industry's leading solution" carries less epistemic weight than a trade publication review saying the same thing. Independence matters.

The Loganix zero-citation study tested 100 queries specifically looking at whether wire domains appeared in AI answers. They didn't, not once. This wasn't a statistical blip. It reflects a structural reality about how AI citation selection works.

When Press Releases DO Drive AI Citations

The Muck Rack finding, a 5× increase in PR-linked AI citations, isn't contradicting the Loganix data. It's measuring something different. The citations growing aren't citations of the wire post. They're citations of the earned media coverage that wire distribution triggered.

When a press release announcing original research gets picked up by TechCrunch, The Verge, or VentureBeat, those secondary articles become citeable. When a funding announcement lands in Business Insider or Forbes, those articles carry domain authority and editorial credibility. When a product launch is covered by a respected industry trade publication, that coverage becomes an authoritative source AI engines will happily cite.

The Muck Rack earned media research found that earned media accounts for 25% of all large language model citations, a remarkable figure that underscores the stakes. PR teams that understand this dynamic are sitting on enormous GEO leverage. PR teams that don't are investing in wire distribution and wondering why their AI visibility isn't moving.

Several press release types have demonstrated higher earned media pickup rates, which correlates with higher downstream AI citation rates:

• Data-driven announcements: Original research, proprietary surveys, new datasets. These give journalists something to write about beyond the announcement itself. A "State of X" study gets covered; a product update announcement usually doesn't.
• Funding and milestone announcements: Seed, Series A, Series B rounds reliably attract tech press coverage on reputable domains. The press release triggers Crunchbase entries and TechCrunch articles that become AI-cited sources.
• Product releases with genuine news value: Launches that solve a recognized problem, or that represent a meaningful category advancement, attract trade press coverage. The key word is "genuine", AI engines are particularly unimpressed by self-reported innovation.
• Partnership and integration announcements: Especially those involving larger, well-recognized brands. The coverage benefits from the partner's brand authority in addition to your own.

Which AI Platforms Are More Likely to Cite PR Wire Content

Not all AI engines treat press release content equally, and the differences are significant enough to inform your distribution strategy.

Perplexity is the most likely to surface wire content directly. Perplexity's architecture is retrieval-augmented: it searches the live web when generating answers and pulls from news wires as part of its real-time sourcing. This means a freshly published press release on PR Newswire has a short window, typically 24–72 hours after publication, where it may appear as a direct Perplexity citation before the earned media coverage materializes. Understanding how Perplexity picks sources is especially useful for timing your PR distribution strategy.

ChatGPT is significantly less likely to cite wire domains directly. Its knowledge comes primarily from training data rather than real-time retrieval, and wire domains don't appear to be heavily weighted in its citation patterns. For ChatGPT visibility, the earned media coverage path is the only reliable route.

Gemini occupies a middle ground, with its real-time Google Search integration meaning that well-indexed news coverage can surface relatively quickly. Google's own news indexing pipeline gives Gemini access to earned media coverage faster than ChatGPT's training update cycle allows.

The ALM Corp press release citation data provides useful per-platform breakdowns on which content types get cited where. Their research reinforces the pattern: direct wire citations are a Perplexity phenomenon; earned coverage citations span all platforms.

The Promotional Content Problem

There's a subtler issue with press release content that even sophisticated PR teams miss. It's not just that wire domains lack authority, it's that the writing style of most press releases actively signals "skip this source" to AI models.

Consider the typical press release structure: opening quote from CEO, superlative product description, boilerplate company description, contact information. None of this is informational in the way AI models prize. Compare it to what AI models do cite: articles that explain how something works, compare options, cite independent data, provide context beyond the announcement itself.

The practical implication: if you want your press release content to contribute to AI visibility, it needs to contain genuinely informational material that journalists can excerpt and that secondary coverage can amplify. A press release that includes original data, clear technical explanations, or substantive expert commentary gives journalists more to work with, and produces better earned media that AI engines want to cite.

A Practical Checklist for PR with AI Citation Potential

Writing press releases that generate AI-citeable earned media requires deliberate structural choices. This checklist addresses the most common failure points:

1. Lead with data, not announcement framing. "Company X releases new feature" generates no coverage. "Company X data shows 73% of enterprise teams fail to X" generates coverage, and the data point becomes a citeable fact in the articles that follow.
2. Include at least one original, citable statistic. This is the single highest-leverage addition to any press release. Journalists need numbers. AI models cite numbers. One proprietary data point can generate coverage that propagates through dozens of articles.
3. Write the press release in inverted pyramid style. Most important information first. AI models processing retrieved content weight earlier text more heavily. Your key facts should be in the first paragraph, not buried under CEO quotes.
4. Target earned media, not wire visibility. Your distribution list matters more than the wire service. Direct pitches to relevant trade journalists outperform wire blasts for earned coverage pickup and downstream AI citation.
5. Build follow-on assets for journalists. A press release alone is rarely enough. Accompany it with a data sheet, a longer research report, or an expert quote document. The more usable material you give journalists, the more substantive their coverage, and the more AI-citeable it becomes.
6. Monitor which publications cover the story. Track whether the coverage lands on high-authority domains. A single TechCrunch article carries more AI citation weight than 50 small-domain pickups. Quality of earned coverage matters more than volume.
7. Follow up with SEO-optimized owned content. Publish a blog post or research page on your own domain that expands on the press release data. This gives AI engines an owned-domain source to cite, and it reinforces the topical authority signals that make your wire-distributed PR more credible by association.

The Amplification Loop: PR as the Trigger

The most useful mental model for AI-era PR is the amplification loop. Press releases, at their best, trigger a cascade: wire distribution creates initial awareness and Perplexity indexing, earned media coverage creates authoritative citeable sources, those sources build your brand's entity recognition across AI training data, and that entity recognition increases the probability of unprompted AI mentions in relevant queries.

This loop means that a single well-executed PR campaign, one that generates substantive coverage in five or ten authoritative publications, can have compounding effects on AI visibility for months or years. The earned media articles don't disappear. They remain indexed, they continue to be crawled, and they continue to accumulate as signals of your brand's legitimacy in AI knowledge representations.

Conversely, a press release that generates no earned coverage does almost nothing for AI visibility, regardless of how well it's written, which wire service distributes it, or how many times it gets indexed. The distribution channel is a means to an end. The end is substantive independent coverage.

Measuring PR's Contribution to AI Visibility

Most PR teams measure success by media placements and readership. For AI visibility purposes, the more relevant metrics are: which publications covered the story, what domain authority those publications carry, whether the coverage includes citable facts and data points, and whether AI engines begin citing those articles in relevant queries. This requires tracking AI citations over time with the granularity to connect earned media events to changes in citation patterns.

The relationship between a press release and its AI visibility contribution is indirect and lagged, coverage takes time to propagate through AI systems, but it is measurable. Brands that run structured tracking around their PR campaigns increasingly find that the highest-authority earned media placements drive measurable increases in AI mention rates within 30–90 days of publication.

Understanding whether your brand appears in AI answers, and whether your earned media is actually driving those appearances, is exactly what BabyPenguin is built for. It tracks brand mentions, citations, and source appearances across ChatGPT, Gemini, and Grok, giving PR and content teams the data they need to connect their distribution activity to real AI visibility outcomes.