Ask a marketing team why they still send releases over the wire and you’ll usually hear about reach: hundreds of pickup sites, a distribution count in the hundreds of thousands, maybe a placement on a well-known outlet’s press release page. None of that reach translates the way most teams assume once the question shifts from “who will see this” to “will ChatGPT ever mention it.”
The short answer is that large language models rarely draw on a press release directly. Wire pickups are treated as duplicate, low-authority text and are mostly filtered out before an answer gets generated. That doesn’t mean distribution is worthless. It means the value shows up at a different stage of the process, one that most companies never measure.
This article looks at what recent citation research actually shows, why AI systems treat press releases the way they do, and where the real opportunity sits for a company trying to appear in AI-generated answers.
What The AI Citation Data Actually Shows
Several research groups have now published large-scale audits of what ChatGPT, Claude, and Gemini actually cite when they answer questions. The results are consistent across studies and industries.
Earned Media Dominates The Numbers
Muck Rack’s Generative Pulse research, now in its third edition, analyzed more than 25 million links cited across ChatGPT, Claude, and Gemini in 17 industries. Earned media accounted for the large majority of citations, while press releases made up less than 2 percent. Separate tracking from Meltwater, covering roughly 5.35 million citations across eight major AI platforms, found press release citations actually falling within a single quarter of 2026, even as overall citation volume from earned and news media grew.
Why Press Release Text Gets Filtered Out
Part of the explanation is structural. A release distributed through a major wire service often appears on hundreds of near-identical pickup pages. AI retrieval systems tend to treat these as duplicate content and collapse them down to a single, low-priority source, or skip them entirely in favor of a page with more unique signal. The release exists online in enormous volume, but that volume doesn’t translate into citation weight.
How AI Models Decide Which Sources To Trust
Citation frequency isn’t random. Researchers studying how language models select sources have found a consistent pattern in what earns a mention.
Outlet Authority Matters More Than Content
A joint study from researchers at Renmin University of China, the National University of Singapore, and the Chinese Academy of Sciences examined how models choose sources and found that the reputation of the outlet, not the content of the piece itself, was the primary driver of citation selection. This authority hierarchy appears to have been shaped during model training, which draws heavily on journalism and inherits its existing sourcing norms.
A Narrow Trust Tier, Not The Whole Web
A separate academic analysis of more than 366,000 citations across tens of thousands of AI conversations found that news citations concentrate heavily among a small number of outlets. In practice, a company isn’t competing against every indexed page on the internet. It’s competing for attention from a fairly narrow tier of publications the models have learned to trust.
The Real Value Chain From Release To AI Citation
None of this means press releases have no role. It means the release is an input into a longer process, not the finished product an AI model will eventually cite.
Distribution Creates Reach, Not Authority
A wire release puts a company’s news in front of a large number of journalists, editors, and databases at once. That reach is real and can be valuable, particularly for companies without existing media relationships. But reach and citation authority are different things, and buying more of the first doesn’t automatically produce more of the second.
Original Coverage Is Where The Value Transfers
Research from Presenc AI illustrates the mechanism well. A release sent to a wire service that produces 200 identical pickup pages and zero original articles adds almost nothing to AI visibility. The same release, if it prompts five independent journalists to write their own articles referencing the company, can generate several new high-authority citation sources within a month or two. The release didn’t get cited. The coverage it generated did.
Why This Changes What “Success” Should Look Like
Most PR reporting still centers on pickup counts and total distribution reach. Based on how AI citation actually works, a more useful metric is how many independent editorial articles a release generated, and at how many distinct, reputable outlets. That number is usually far smaller than the pickup count, and it’s the one that predicts AI visibility.
What This Means For Companies Choosing A Distribution Strategy
None of this argues against using a distribution service. It argues for choosing one, and using it, with a clearer goal in mind.
When Wire Distribution Still Earns Its Cost
Companies without an existing network of journalist relationships often find wire distribution to be the fastest way to get in front of reporters who might otherwise never see their news. For early-stage or lesser-known brands, that reach can be the cheapest available path to the kind of original coverage that eventually earns AI citations.
Questions Worth Asking Before You Buy
Before committing budget to a distribution package, it’s worth asking which specific outlets and journalists the service actually reaches, whether the release is written in a way that gives a reporter something worth writing about independently, and how pickup will be tracked separately from any original coverage it generates. A service that can answer all three is solving the right problem. One that can only point to a pickup count is optimizing for a number that AI models have already learned to ignore.
Frequently Asked Questions
Does ChatGPT ever cite press releases directly?
Occasionally, but rarely. Research from Muck Rack found press releases account for under 2 percent of citations across ChatGPT, Claude, and Gemini, while earned media accounts for the large majority.
If press releases aren’t cited, why distribute them at all?
Distribution’s main value is reaching journalists and prompting original coverage. That coverage, not the release itself, is what AI models tend to cite.
What matters more than pickup count for AI visibility?
The number of independent editorial articles a release generates at reputable outlets matters more than the total number of syndicated pickup pages.
Which companies benefit most from wire distribution?
Early-stage or lesser-known companies without existing journalist relationships tend to see the most value, since the wire often serves as their cheapest path to original coverage.
The pickup count on a distribution report was never a great proxy for reach, and it’s an even worse proxy for AI visibility. The companies getting real value from press releases in 2026 have mostly stopped treating the wire as the finish line. They’re using it as the first step in a chain that, if it works, ends with a journalist writing something original, and an AI model deciding that journalist’s work is worth citing.