PRESS RELEASES × AI VISIBILITY

Every input that gets your brand cited by AI is, at root, somebody vouching for you. A journalist, a reference site, a backlink from a publisher who chose to point at you. A press release is the one exception: it is the reference you get to write yourself and pay to syndicate. That is what makes it tempting, and it is exactly why the engines trust it least. Run the flag up your own mast and you have raised one voice; the engines are counting how many others answer.

THE SHORT VERSION

This page is part of our AI visibility hub. Below: what a release actually is to an AI engine, the narrow window where the format is worth running, and how to use it without overpaying for the other 99%. For the wider earned-media play it sits inside, see digital PR for GEO.

~1%

of all AI citations come from press releases

Muck Rack, 2026

~5x

growth in PR AI citations since July 2025

Muck Rack, 2026

~84%

of AI citations come from earned media you did not author

Muck Rack, May 2026

Strip away the sales pitch and a press release is a structured, dated, attributed document that you wrote about yourself and paid to distribute. The first three properties are genuinely useful: a retrieval engine can lift a clean, time-stamped claim from it without much guesswork, which is more than it can say for most marketing pages. That is the real, narrow value of the format.

The fourth property is the catch. The engines do not just read a source, they weigh how many independent sources say the same thing. A press release is a single voice, and it is the voice with the most obvious motive. So it counts, but it counts as one self-authored data point, not as the chorus of unrelated coverage that actually moves an engine. An engine can tell a chorus from an echo.

That is why the headline number on this channel is so small. Across the web the models read, press releases account for only about 1% of all AI citations, even though the channel has grown roughly fivefold since July 2025. Real, rising, and tiny, all at once. The 1% is the figure most distribution vendors leave out, and it is the one that should set your budget.

Send only genuine news

A real launch, funding round, partnership, or original dataset gives an engine something credible to lift. A release about nothing is a flag with no wind. A steady drip of manufactured releases actively trains both engines and editors to discount you.

STEP 2

A direct answer in the first sentence or two

Put the answer immediately under the heading, before any elaboration. This is the single habit that does the most for you here, and there is a reason it matters more now than it did for snippets. Roughly 44% of the citations one large study of ChatGPT pulled came from the first third of the page. The engine reads the top and rarely digs. Bury the answer three paragraphs down and a snippet might still find it; an answer engine usually will not.

Treat it as corroboration, never the foundation

A release works when it confirms a story your editorial coverage and your own pages already tell. Build the story first; let the release second it. On its own it is a single point.

A release earns its citation when it stops talking to itself. Put genuine news on a credible distribution service and your own dated statement sits beside coverage other people are going to write anyway. Now it is no longer one voice: a journalist’s account, a publisher’s link, and your syndicated release describe the same event in the same window. That alignment is what an engine reads as real.

The direction is consistent with everything else the engines do: getting genuine news onto credible, crawled outlets corroborates your story, and corroboration is the currency. The release alone underperforms the most, because it is the smallest and the only self-authored slice.

Picture three inputs the engines read as independent corroboration: editorial placements on real publications, authoritative brand mentions and links, and press distribution for genuine news. The first two do most of the work. The release adds a dated, syndicated, citable signal that seconds the same story. Together they build the consensus a model surfaces; any one alone underperforms, and the release alone underperforms the most.

The release treadmill & spray-and-pray

Sending something out every week to stay visible erodes the credibility that made the format work. Cheap services that blast a release across hundreds of low-quality sites add close to nothing, because the engines weigh the credibility of the carrying outlet, not the raw count. A copy on a site no model trusts is the same voice in an empty room.

Treating PR as the whole plan

This is the expensive mistake. Because releases are about 1% of AI citations, a program built mostly on them leaves the other 99% on the table. PR is a supporting instrument. If it is your lead instrument, the arrangement is wrong.