AI for PR Pros: What's Actually Useful

PR reached AI saturation before most of marketing and then stopped: Muck Rack’s State of AI in PR, surveying 564 PR professionals, has adoption at 76% and “mostly unchanged from last year.” The tooling question is settled. What is not settled is the craft question, and PR has the sharpest version of it anywhere in my role-by-role series: the entire job runs on a relationship with people - journalists - who can tell when you automated them, and say so.

My organizing rule for this seat: AI on the reading and the drafting, never on the relationship. PR’s inputs are endless reading and its outputs are trust, and only one of those scales.

The reading half: monitoring and research, handed over gladly

The unambiguous win is the reading. Coverage monitoring, competitor announcement tracking, journalist-beat research, transcript synthesis, the daily clip review - this is high-volume reading that a model summarizes tirelessly, and it maps to what PR pros already do with AI: in the Muck Rack data, 86% use it for editing and refinement and 76% for research and insights, the two least glamorous line items in the report. My version of the daily practice: coverage and competitor mentions in, a structured brief out - what was said, by whom, sentiment called conservatively, and anything that needs a human decision flagged rather than answered. The model reads everything so I read the five things that matter.

Two fences on the reading, both learned from the general hallucination problem. Summaries cite their source articles so claims can be checked in one click, and sentiment gets treated as a first pass, not a verdict - a model reading irony in trade press is guessing, and crisis-adjacent coverage gets human eyes regardless of what the summary says.

The pitch paradox, and what it actually teaches

Here is the data pair that defines modern media relations. Muck Rack’s State of Journalism has 82% of journalists using at least one AI tool themselves, up from 77%. And in Cision’s State of the Media, 53% of journalists oppose receiving AI-generated pitches, against 21% in favor. The people rejecting your AI pitch are using AI to do their own jobs the same morning.

I do not read that as hypocrisy. I read it as a precise statement of what a pitch is for: a pitch is evidence that a specific human matched a specific story to a specific journalist’s beat. AI-generated pitches fail because they are evidence of the opposite - that no one did that work - and journalists, drowning in volume, use exactly that signal to triage. The practical rule that falls out: AI drafts the pitch’s raw material (the summary of the announcement, the data points, the boilerplate), and the first two sentences - why this journalist, why now - get written by the human who actually knows. If you cannot write those two sentences, the problem is the targeting, not the drafting, and no model fixes targeting you did not do.

Releases, statements and the fabrication line

Press releases, media kits, Q&A docs and briefing materials are structured, convention-bound writing - good AI drafting territory, and the 74% of PR pros using AI for writing are mostly here. The fence that matters more in PR than anywhere else is fabrication, because PR’s product is placed trust: a hallucinated statistic in a blog post embarrasses you, but one in a press release gets printed under a journalist’s byline, and that journalist remembers who burned them permanently. Every factual claim in anything sent to media goes through a verification pass with the receipts attached, and quotes attributed to executives are approved by the executive, verbatim, every time. The adjacent line I would treat as absolute: no synthetic people. Fake spokespeople, invented experts and AI-generated commentators offered to real journalists is the fastest reputation suicide available in this profession, and the brand safety rules apply to the agency side double.

Crisis: speed from the machine, judgment from the room

Crisis work tempts teams toward AI for the worst reason - speed under pressure. The usable version is narrow: AI monitors the spread and summarizes it in real time, drafts holding-statement variants from your approved facts, and pressure-tests your draft by arguing the hostile reading of every sentence. The decision layer - what to concede, when to speak, who speaks - stays in the room, because a crisis is precisely the moment your judgment is the product. My rule: in a crisis the model reads faster so humans can think longer, and nothing model-drafted ships without the most senior review available.

The new measurement surface: what AI says about your clients

PR’s measurement question - who saw it, what did it change - has a new wrinkle worth owning early: AI assistants now answer questions about your brand or client, and earned media is a large part of what they draw on. The monthly audit from my brand manager playbook - the fifteen highest-stakes questions run through the major assistants, answers logged, errors turned into correction targets - is PR work in all but name, and the evidence that brand mentions correlate with AI visibility makes coverage the input. For comms teams this is a genuinely new deliverable to sell: what the machines say when someone asks about the client, and how to change it.

What stays yours

Three things I would not automate in this seat, structurally. The relationships - a journalist’s trust is built on you being reliably useful and honest over years, and no volume of AI-personalized outreach substitutes for the phone call that says “this one isn’t for you, but Thursday’s is.” The judgment about attention - knowing what is actually a story, what the room will ask, and what should never be said is the compressed experience clients pay for. The accountability - when coverage goes wrong, “the AI drafted it” is not a sentence a PR professional survives saying. Half the profession now has formal AI policies - 51% in Muck Rack’s data, up from 21% in 2024 - and the good ones all encode the same idea: the tools assist, the practitioner answers for it.

FAQ

What should a PR pro automate first? Monitoring and coverage synthesis - the highest-volume reading in the role, zero relationship risk, and the output improves every other decision you make the same day. Pitch personalization is the last thing to automate, for the reasons the Cision numbers make plain.

Will journalists know if my pitch is AI-written? Assume yes. They use the tools daily themselves, they see hundreds of pitches, and the generic-pitch pattern is exactly what they triage against. The safer frame: a pitch’s value is proof of human targeting, so automate everything about the pitch except the proof.

Can AI write a press release? It can draft one competently from a fact sheet, and that is a real time saving. The release still needs the verification pass, approved quotes, and a human read for the claim you would not want challenged on the record - the drafting was never the expensive part of a release, the being-right was.

Will AI replace PR professionals? It compresses the production layer - monitoring summaries, first drafts, coverage reports - and leaves the parts that were always the job: relationships, story judgment and standing behind what ships. A PR career built on volume outreach has real exposure. One built on being trusted by fifty journalists does not, and the tools just freed hours to deepen exactly that.

This post is one of the role-by-role playbooks - the nearest neighbors are AI for brand managers on the reputation surface and AI for social media managers on the adjacent channel. The verification discipline lives in the fact-checking prompts, and the wider practice starts at the complete guide to AI marketing.