PR teams should tell journalists when AI produced images, video, audio, quotes or research data, and need no disclosure for AI-assisted drafting that a human edited and owns. Muck Rack's State of Journalism 2026 found that 82% of journalists use AI tools and 88% disregard pitches that miss their beat. A pitch that also hides AI-made material loses its reader and the sender's credibility.
Last updated October 9, 2026. Part of EPR's coverage of AI in PR operations.
How many journalists and PR professionals use AI?
Both sides of the pitch now use AI as a routine tool, so the question is disclosure, not adoption. The table shows the latest Muck Rack survey figures for each group.
| Group | Finding | Source |
| Journalists | 82% use at least one AI tool, up from 77% the year before | Muck Rack State of Journalism 2026 (897 responses, surveyed January 30 to March 2, 2026) |
| Journalists | 86% say at least some of their stories originate from PR pitches; 88% disregard pitches that miss their beat | Same report |
| PR professionals | 76% use generative AI in their workflow; 51% work somewhere with an AI use policy; 12% of AI users use AI agents | Muck Rack State of AI in PR 2026 (January 2026) |
The 51% policy figure means close to half of PR professionals do not report an AI use policy at their workplace. That gap is where most disclosure failures start.
What do newsrooms expect from outside material?
Newsrooms expect outside material to be vetted, and some state that they screen incoming content for AI. The Associated Press, in its standards around generative AI, treats any generative AI output as unvetted source material. It does not allow generative AI to add or remove elements in photos, video or audio, and it allows an AI-generated illustration only when the illustration is the subject of a story and is labeled as such in the caption.
The same AP page encourages journalists to make sure material arriving from other sources is free of AI-generated content. For a publicist, that means an unlabeled AI image or a synthetic statistic can get a pitch rejected, or get a published story corrected later.
For EU audiences, disclosure is also a legal duty for deepfakes and for AI-written text on matters of public interest, as of August 2, 2026. EPR covers the rules in its EU AI Act Article 50 checklist.
What should PR teams disclose to journalists?
PR teams should disclose AI use based on what the AI produced, not on whether AI touched the workflow at all. The table sets out EPR's recommended defaults.
| AI use | Disclose? | What to do |
| Drafting help, then human edit and review | No | Treat the pitch as the publicist's own work. The human is responsible for every fact, as with spell-check or grammar tools |
| Generated or edited images, video or audio | Yes, always | Label the asset at the point of distribution. A product image, an AI-edited headshot or a cloned audio clip all need a label |
| Quotes | Do not use AI quotes | Take quotes from statements an identifiable person made. Synthesizing a quote and getting sign-off afterward creates a documentation problem that grows |
| Survey data or research simulated by an LLM | Yes, upfront | Label the output as modeled, never as audience data. Reporters who find simulated data after the fact treat it as a red flag |
| Machine translations and summaries of long documents | Yes, if material | Tell the reporter, so their fact-checking allows for the added uncertainty |
What does a failed disclosure cost a PR professional?
A failed disclosure costs the sender future access, because a reporter who finds undisclosed AI material after a story runs has reason to doubt the next pitch from that source. The upside of skipping disclosure is a few seconds saved per pitch. The downside is a pitching relationship that may not recover.
This is the same mechanism behind PR's trust problem with itself: each shortcut that gets caught lowers the benefit of the doubt for the whole industry. The PRSA Code of Ethics puts honesty and disclosure of information among its core provisions, and applying them to AI use is straightforward.
What belongs in a written AI use policy?
A written AI use policy should state who is responsible for accuracy, what must be disclosed, and what is banned. Four rules cover most cases.
| Situation | Rule |
| AI drafts text that a person edits and reviews | No disclosure is required. The named person is responsible for accuracy |
| AI generates images, video or audio for external distribution | Disclose at the point of distribution |
| AI generates quotes or content that could pass as primary research | Do not use it outside the team until a human source is verified |
| A use is unclear | Disclose |
The decision for a communications team is to adopt these four rules in writing this month, and to name one person who signs off on every image, quote and statistic that leaves the building. EPR's Generative Engine Optimization canon sets out how AI-native communications teams operate in general, and AI-Generated Ads: The Disclosure Crisis Coming in 2027 covers the advertising side. For the audience context, see EPR's Reuters Digital News Report 2026 analysis.