Bridgehead Communications

Insight · Corporate

Press releases and AI search: what communicators believe and what AI tools cite

By William Walter

Managing Director, Bridgehead Communications

Published 6 min read

PR Newswire's 2026 survey of 1,016 communicators says releases help AI search. We put 30 UK health and social care questions to three AI tools, and none of the 90 answers cited a wire service.

Press releases and AI search: what communicators believe and what AI tools cite

Key findings

  • None of the 90 AI answers in our UK health and social care test cited a wire service. The 738 citations included none from PR Newswire, Business Wire or GlobeNewswire.
  • 27% of those citations (201 of 738) went to government, NHS and regulator sites. gov.uk alone appeared in 31 of the 90 answers.
  • In PR Newswire's survey of 1,016 communicators, 76% measure media pickup and only 34% track visibility or citations in AI-powered search.
  • 47% of survey respondents are confident AI tools cite their releases, and 11% do not measure it at all.
  • 36% of EMEIA respondents say AI search has changed how they optimise releases, against 57% in APAC.

PR Newswire published its State of the Press Release 2026 report this autumn. It draws on a survey of 1,016 communications and marketing professionals in more than 30 countries, run between 9 September and 12 October 2026. We read it with one question in mind: are communicators changing how they write releases because of AI search, and can they show it is working?

Many are changing. Few can show it. The report also says large language models treat wire releases as authoritative and cite them often, and gives no data for that. We tested it ourselves.

What AI tools cite on UK health and social care questions

On 18 November 2026 we put 30 questions to ChatGPT, Perplexity and Gemini, each with web search switched on. Ten were news questions (for example, "CQC announces changes to inspection approach"), ten were commercial ("best digital care planning systems for care homes UK") and ten were informational ("how is social care funded in England"). That gave 90 answers. We recorded every URL they cited and checked each against 23 wire and release-distribution domains, including PR Newswire, Business Wire, GlobeNewswire, PR Web and Pressat.

The answers carried 738 citations. None came from a wire service. The result held for every engine and for all three question types, including the news questions, where a wire release would be most likely to turn up.

Official sources dominated. Government, NHS and regulator sites took 201 of the 738 citations (27%). gov.uk appeared in 31 of the 90 answers, england.nhs.uk in 13 and cqc.org.uk in 9. Trade titles such as Digital Health and HTN were cited regularly. On product questions the tools cited company websites: The Access Group's site appeared in 6 answers and Birdie's in 6. Two of the cited URLs on organisations' own sites looked like releases or news items, and both came from Perplexity on news questions.

Six ChatGPT answers and two Gemini answers cited nothing. All eight were informational questions.

How far to trust the test

It is a small, exploratory sample. We chose the questions, and we ran each one once, so a rerun would shift individual citations. We used the APIs, which can differ from the consumer apps. Perplexity's count is of retrieved sources and not only those it named in the answer. Gemini's domains come from its grounding data. A wire release can also reach an AI answer indirectly, because trade titles republish releases, and we did not trace that. Our test shows what the tools cite directly. It does not show that a wire release has no effect.

Within those limits the pattern is clear enough to act on. Across 738 citations on 30 questions, the wire services were absent.

Teams are changing their releases faster than they can measure the effect

Asked how they judge a release, 76% of survey respondents said media pickup. Half named website traffic or engagement and 38% named search rankings. Only 34% track visibility or citations in AI-powered search, and 25% track leads or other business outcomes. A further 10% do not measure release success at all.

Confidence runs ahead of measurement. 13% of respondents are very confident that AI tools find or cite their releases and 34% are somewhat confident, which makes 47% confident. Another 11% say they do not measure it. Our test gives that confidence little to rest on, at least for wire distribution in UK health and social care. The report itself notes that traditional search rankings say little about how often AI-generated answers mention a release.

PR Newswire reports its results by region, and its EMEIA group covers more than the UK, so none of these figures is UK-only. 36% of EMEIA respondents say AI search has changed how they optimise releases, against 41% in North America and 57% in APAC. 41% of EMEIA respondents say they have not changed how they use releases, against 14% in APAC. 15% of EMEIA respondents track new AI-related metrics, such as AI citation volume, against 41% in APAC.

What respondents say they have changed

Asked what they have done to make releases more likely to be cited by AI tools, respondents most often named clearer, more descriptive headlines (57%). Relevant keywords came next (42%), then answering key questions directly in the release (41%) and clear sections and headings (40%). Conversational phrasing and statistics or original research each came in at 32%. 30% link to supporting content on their own website, and 17% have done nothing.

The figure for statistics stands out. Headlines and headings are cheap to change. Original data takes work, and it is the part of a release that a journalist has a reason to quote.

AI drafting is now the norm

70% of respondents globally use generative AI on some part of their release content. In North America the figure has risen from 26% two years ago to 65%. It stands at 73% in EMEIA and 80% in APAC, and 12% say they never will. The most common use is a first draft or outline.

When most communicators draft with the same tools, the draft is the least distinctive part of a release. A named person saying something specific, and a figure nobody else has, are what is left. That is our reading, not the report's, but the respondents point the same way: they ranked the executive pull quote second for impact, behind the headline, and put the key takeaways section third.

Coverage is still how most people judge success

76% still measure media pickup. Respondents credit wire distribution with raising brand or product visibility (60%) and with reinforcing their authority in their industry (45%). 39% credit it with SEO, AEO, GEO or link-building benefits. PR Newswire notes that responses were higher than last year in most categories.

How to read the survey

PR Newswire sells wire distribution, and the report supports that business. Three limits matter. The benefits are what respondents believe, not outcomes anyone measured. The report's claim that large language models cite wire releases often comes with no citation data, and our test found none in 90 answers. The regional figures also group the UK with other countries.

What we take from it

Wire distribution can be worth buying for reach and for coverage. On our evidence, AI visibility is not a reason to buy it. In UK health and social care the tools cite regulators, government, trade titles and company websites, and that is where a business needs to appear.

A release written for people already suits machines. A clear headline, a short summary of the news and headed sections are what a journalist wants, and the survey says most teams already do them. Put the same material, with its figures, on your own site, where a tool can retrieve it.

The gap is measurement. Reporting that sets media coverage beside AI citations and website traffic is uncommon in this survey, and 34% tracking the AI side is a low base. A fixed set of questions put to the main tools each month is cheap to run, and it replaces confidence with a number. Original data is the other gap. Few teams include it, and it is the part of a release that people and AI tools are most likely to quote.

Sources and method

What is the State of the Press Release report?

It is an annual survey run by PR Newswire, a press release distribution service. The 2026 edition covers 1,016 communications and marketing professionals in more than 30 countries, surveyed between 9 September and 12 October 2026.

Do press releases distributed on a wire service appear in AI-generated answers?

Not in our test. We put 30 UK health and social care questions to ChatGPT, Perplexity and Gemini on 18 November 2026. The 90 answers carried 738 citations, and none came from a wire service. The sample is small and the questions were ours, so it shows a pattern in this sector and does not prove wire releases are never cited.

What did the AI tools cite instead?

Official sources first. Government, NHS and regulator sites took 27% of citations, and gov.uk appeared in 31 of 90 answers. Trade titles such as Digital Health and HTN came next, along with company websites on product questions.

How did you run the test?

We asked each question once through the OpenAI, Perplexity and Google Gemini APIs with web search switched on (UK location for ChatGPT). The questions were split evenly across news, commercial and informational intent. We recorded every cited URL and checked it against a list of 23 wire and release-distribution domains. Perplexity's figure counts sources it retrieved, and Gemini's domains are read from its grounding data.

How can a UK team measure whether a release is cited by AI tools?

Put a fixed set of questions to the main AI tools each month and record which sources they name. Track that result next to media pickup and website traffic so the three can be compared.

Stay in touch

Get the Corporate newsletter

Speak with us

One conversation tells you
if we're the right fit.

Thirty minutes with a partner. We'll tell you whether we can help, what the work would look like, and what it's likely to cost, before either of us commits to anything.