Generative AI adoption in public relations has plateaued, per Muck Rack's State of AI in PR 2026 report, released January 21, 2026. Based on a survey of more than 500 PR professionals, the report finds 76% now use generative AI in their work — roughly unchanged from the previous year — leading Muck Rack to describe adoption as having "leveled off."
The numbers behind the plateau
- 76% of PR pros use generative AI — the ceiling has effectively been reached for individual tool usage.
- 43% of companies now offer AI training to PR staff, an increase over the past year, but still under half.
- The report tracks growing experimentation with AI agents — autonomous multi-step tools — and with governance: who is allowed to use what, on which tasks, with what disclosure.
What training should actually cover
The 43% training figure is the number to interrogate, because "training" is a vague word. A one-hour onboarding webinar checks the survey box but does not change behavior. What the failure modes demand is narrower and more concrete: verification habits (every fact, quote and citation in AI-assisted output gets checked against a primary source), confidentiality rules (what client information may never be pasted into a consumer tool, and which tools are approved at all), and disclosure norms (when an output's AI provenance is stated to a client or journalist, and in what form). Each of these is a policy paragraph plus a checklist, not a course curriculum.
The organizations that moved early treat it as a quality function, not an IT function. The person who owns pitch accuracy owns AI accuracy, because from the recipient's point of view there is no difference — a fabricated statistic damages credibility whether a human or a model invented it. Teams that formalized this when adoption was climbing are now operating normally; teams that wait are accumulating quiet risk inside every un-reviewed draft.
Why the plateau matters
For two years, the industry's AI story was adoption — who had started using the tools. That story is over: nearly everyone has. The new dividing line in the data runs between organizations with structured approaches and everyone else. Training coverage at 43% means a majority of PR professionals are using powerful language tools with no formal guidance on accuracy, confidentiality or disclosure. The predictable failure modes — fabricated citations in pitches, client data pasted into consumer AI tools, undisclosed AI-drafted outputs — are process failures, not tool failures.
There is also a competitive-quality reading. When 76% of the market uses the same models, raw tool access confers no advantage; the outputs converge. What still differentiates is what the tools are fed — proprietary data, original research, real client work — and the editorial judgment to know when a draft is confidently wrong.
Related stories: 82% of Journalists Now Use AI — Muck Rack's State of Journalism 2026 Maps the Newsroom · PRWeek's Agency Business Report 2026: Growth Slows to 3% and the Market Splits in Two.
The buyer's side of the survey
Read as a procurement document rather than an industry snapshot, the report is a due-diligence template. If you hire agencies or freelancers, the plateau means you can no longer distinguish vendors by whether they use AI — everyone does. The distinguishing questions have moved one level up: What is your policy on AI use in client work? Who reviews AI-assisted output before it reaches a journalist or a client? What happens to client data inside your tools? Which outputs, if any, do you disclose as AI-assisted?
Agencies with real answers — a written policy, named reviewers, a data-handling rule — can answer those questions in a sentence each. Agencies that have not thought about it will answer with enthusiasm about the tools instead. That gap is now the fastest available signal of operational maturity, and it takes one meeting to detect.
A practical verification layer: ask for the policy in writing and check whether it mentions error rates, not just capabilities. Documents that only celebrate efficiency were written for marketing; documents that assign responsibility for accuracy were written for operations.
The agent era arrives quietly
The most forward-looking section of the report covers agentic AI: systems that execute multi-step workflows — build the list, draft the pitch, schedule the follow-up — rather than answer single prompts. Most practitioners surveyed remain early in experimentation, and the governance questions multiply with autonomy. An errant chatbot draft wastes an afternoon; an errant agent with media-list access can email a thousand journalists. The report's implicit advice is the industry's near-term agenda: extend existing media-relations quality controls to AI-driven workflows before scale makes the mistake expensive.
The takeaway for teams
If your organization has not formalized AI use in PR — a written policy, basic training, human review on anything client-facing — the survey says you are now below the median, not cutting-edge. And if you have, the next differentiator is already visible on the report's horizon: measured, supervised use of agents on the mechanical layer of the job, freeing human hours for the judgment layer that no model supplies.
A working checklist
For teams closing the gap between the 76% and the 43%, the sequence is short:
- Write the one-page policy first — approved tools, forbidden inputs, disclosure defaults. Training without a policy teaches opinions.
- Route all client-facing output through a named human reviewer regardless of how it was drafted.
- Log AI-assisted work somewhere countable, so the policy's real usage — not its intended usage — is visible after a quarter.
- Pilot one agent workflow on low-stakes internal work before anything touches a media list.
None of this is advanced, which is the point. The survey's middle tier — the majority using powerful tools with no formal guidance — is not short on intelligence; it is short on a process someone wrote down.
Per Muck Rack State of AI in PR 2026, released January 21, 2026.
For more context, read 82% of Journalists Now Use AI — Muck Rack's State of Journalism 2026 Maps the Newsroom.
For more context, read cision state of the media 2026.
For more context, read eu ai act article 50.