Do press releases help you get cited in AI search?

How often do press releases actually get cited by AI tools?

In buyer-intent "best of" questions, almost never. Loganix found press release domains cited in zero of the 100 category queries it ran in March 2026.

We heard about that zero from a press release. Loganix announced the result on PRNewswire on March 24, 2026, and the announcement sits under a path called /news-releases/. Loganix is an SEO and digital PR agency that now sells an AI press release service built on what it found, so we read the study with that in mind.

The test spread 100 category queries across Perplexity, ChatGPT, and Gemini, covering 10 verticals that included SaaS, finance, legal, and healthcare. None of the answers cited a press release domain, a 0 percent rate. These are the questions a prospect asks before building a vendor shortlist.

We could not find a primary study page, only the release, so we treat the zero as Loganix's reported result and not as a dataset anyone can check.

Why do "/press-releases/" URL paths get filtered out?

Loganix's explanation is that AI systems read that path as self-reported and set the page aside before weighing what it says. Its evidence is a second test that moved the same content between addresses.

In Loganix's eight-path citation test, identical content was placed on eight different URL paths and tracked on ChatGPT, Gemini, Perplexity, and Google AI Overviews. Paths that read as editorial, such as Yahoo Finance's /news/ and Business Insider's /markets/, earned repeat citations. The "/press-releases/" sections at Fortune, Reuters, and wire outlets almost never did, despite comparable domain authority.

That last clause matters if you assumed a strong domain lends its weight to every page on it. Our piece on domain authority and AI citations looks at how much a site's strength counts on its own.

Which types of AI queries do press releases occasionally reach?

Industry-trend questions, in small numbers. Muck Rack's May 2026 data, as AuthorityTech.io reports it, puts a press release citation in about 1 percent of industry-trend answers, roughly 3.5 times the rate for "best of" questions.

AuthorityTech.io's write-up of Muck Rack's data covers more than 25 million links cited by ChatGPT, Claude, and Gemini across 17 industries. Muck Rack, a media database and PR platform, did the classification. The article was updated on September 9, 2026, which makes it the most recent source we used.

The overall share is growing from a low base. Muck Rack's December 2025 report put press releases at about 1 percent of all cited links, up from 0.2 percent in July 2025. That figure counts every kind of question, so it is a separate number from the industry-trend share, even though both land near 1 percent.

Loganix's zero measured something else again. It counted buyer-intent category queries and found no press release citations, while Muck Rack sorted answers by question type and found a thin, real slice in trend questions. A release about a shift in your market has some chance of turning up in an answer about that market, and very little in an answer about which vendor to hire.

What should a press release do differently to be read as editorial rather than promotional?

It should aim at the people who might write about it, because the words alone did not decide the outcome in Loganix's test. Identical text earned citations under editorial paths and almost none under "/press-releases/".

Most guides to writing a release for AI search offer formatting tips with no measured citation rate behind them. The data points elsewhere. In Muck Rack's May 2026 sample, as AuthorityTech.io reports it, 84 percent of citations went to sources brands neither owned nor paid for, and journalism alone made up 27 percent.

So we would write a release for the reporter or trade editor who might cover it, with a finding they can check and quote, and treat the coverage it earns as the part with a chance of being cited. A release is still a fast way to put a fact on the record, and that has value even when no assistant quotes it. Our piece on G2 and Capterra reviews looks at another kind of third-party source.

We run the AI search work for clients who want a team on it, and earning mentions in sources a brand does not own is part of it. That work is slow, and movement usually shows in one to three months.

The PRNewswire release from the top of this post worked that way on us. We learned about Loganix's zero from it, and this page now repeats the finding on a site Loganix neither owns nor paid for.