Field notes4 min read

Press release wires were 272 of 201,185 AI citations, and none were the tracked brand's

When an engine did cite a release, it was usually about a company it had already decided to name, and it preferred the copy on that company's own site.

We checked 201,185 cited pages for press release wires

The sample is every answer Proofsource collected from October 4 to 27, 2026, for an English buyer question that doesn't name the brand, on ChatGPT, Perplexity, Claude and Google AI Overviews: 18,858 answers for 119 brands in 80 workspaces. Those answers cited 201,185 pages.

We counted a citation as a wire page when it pointed at a distribution service such as Business Wire, PR Newswire, GlobeNewswire, EIN Presswire, PRWeb or openPR.

Wire pages got 272 citations, about one in 740

They showed up in 217 of the 18,858 answers (1.2%). By engine, as a share of each engine's citations:

  • Claude: 144 of 62,119 (0.23%).
  • Perplexity: 98 of 69,645 (0.14%).
  • Google AI Overviews: 18 of 33,861 (0.05%).
  • ChatGPT: 12 of 35,560 (0.03%).

Almost half of Claude's wire citations were holiday deal releases from 2020 to 2023

66 of Claude's 144 came from eight Business Wire releases dated 2020 to 2023, all cited on questions about buying a snowboard online. The four whose titles show in the address were Black Friday deal roundups, two of them naming a retailer that a user tracks. A wire page stays up for years, and a name in its title can pull it into an answer long after the sale ended.

Perplexity's wire citations were current. Of the 49 whose date sits in the address, 45 were from 2026.

Not one cited release came from the brand being tracked

None of the 272 wire citations was a release issued by the brand the answer was tracked for, matching on the company name in the release's address. 57 were releases about one of that brand's tracked competitors, and the answer named that competitor in 55 of them.

We read that as support for a name the engine already had. These were established companies it would likely have named anyway, and the release was the page it used to back up a sentence about them.

Companies' own newsroom pages were cited about six times as often

Pages in the press, newsroom or investor section of a competitor's own site were cited 347 times, against 57 for wire copies about those competitors. ChatGPT cited competitors' own newsroom pages 98 times and any wire page 12 times.

Smaller tests point the same way. Loganix ran 100 "best X in Y" questions across ten industries on Perplexity, ChatGPT and Gemini in March 2026 and found no press release domain cited in any of them. Loganix sells press release placement, so that finding cuts against its own product. OtterlyAI reported its own two releases cited within days in September 2025, through copies on Business Insider and Yahoo Finance. That was one company, measured on the questions it tracks for itself.

What we'd do with a press release budget

  • Don't buy wire distribution to get into answers to category questions. If you buy it, buy it for the journalists who might write about you.
  • Publish the full announcement on your own site at an address that won't move, with the facts a buyer would ask about: what it does, who it's for, what it costs, when it shipped.
  • Search the wires for old releases that name you. A five-year-old deal page can still be the thing an engine reads.

How Proofsource handles it

Proofsource records every page each engine cited in each answer, next to the brands it named. You can see whether any engine read your releases, a wire copy of them, or your newsroom, on the questions your buyers ask.

Where this could be wrong

We matched wires by domain. Copies of releases on finance portals are missed: finance.yahoo.com and markets.businessinsider.com got 124 citations between them, many of them release copies. Counting every one would still leave release pages near 0.2% of citations.

Newsroom pages were matched by address, so a few are research firms' or newspapers' press pages, not a company's own.

We don't know how many of the 119 brands issued releases in the period, so zero cited own releases may partly reflect how few existed. A release can also reach an answer indirectly, through the articles journalists write from it or through training data. We only measured what was cited live.

The sample is 24 days, English questions asked from the US, and the categories our users chose to track. One category accounts for most of Claude's wire citations.

Where the outside numbers come from

  1. Zero Press Release Domains Appeared in Loganix's 100-Query AI Category Test · Loganix via PR Newswire
  2. AI Search Engines Love Press Releases · OtterlyAI
  3. Citations · Proofsource
All notes

Questions

Do press releases help my brand get mentioned in ChatGPT and other AI answers?

Not directly, in our October 2026 sample. Wire services got 0.14% of 201,185 citations across four engines, ChatGPT cited a wire page 12 times in 35,560 citations, and no tracked brand's own release was cited.

Which AI engine cites press releases most?

Claude, at 0.23% of its citations, then Perplexity at 0.14%. Almost half of Claude's came from old holiday deal releases on one category's questions.

Where should I publish company news so AI engines can use it?

On your own site. Engines cited pages in competitors' own newsroom and press sections about six times as often as wire copies about the same companies.

Your own numbers beat our best post.

The shortlist in your category is being written right now, whether anyone reads this page or not. A free trial, 25 questions on four engines today and tomorrow, tells you if your name is in it.

200 answers. 25 Prompts/Question. Top AI engines. 2 days. Free.ChatGPTPerplexityClaudeGoogle AI OverviewsGoogle AI ModeGeminiMicrosoft CopilotGrokDeepSeekMeta AI