It's 4.55pm. A client has seen a story break and wants a reactive comment by half five. The coverage report is due Thursday, the new business deck was due yesterday and somebody on the account team is on their third draft of a release about a product that is, with the best will in the world, a slightly better bin. PR has always run on deadlines and adrenaline. AI won't touch the adrenaline, but it can take a decent bite out of the deadlines.
Start PR agency AI training with first drafts of the writing your team produces every week: releases, pitches, social copy and reports. Add verified research and media list support, then a reactive comment workflow with firm client sign-off. Finally, teach the team how AI search engines read and cite earned media, because that is fast becoming part of the job.
Why PR account teams need proper AI training
Most of your team already use it. Muck Rack's State of AI in PR 2026 found 76% of PR professionals use generative AI, and 98% say they always or often edit what it produces. That second number is the reassuring one. The profession knows a first draft isn't a finished draft.
The rules are shakier. CIPR's 2026 State of the Profession report found a quarter of practitioners use AI at work without any formal policy or guidance, and 76% believe AI-generated content risks damaging trust in public relations. Using AI without rules, in an industry whose only real product is trust, is a brave choice.
Training closes the gap between those findings. Almost everybody uses AI. Fewer use it well, and fewer still could explain to a client exactly how they use it and what never goes near it.
- 01Draft fasterReleases, pitches, social.
- 02Research properlyCheck every name and number.
- 03React with sign-offSpeed without regret.
- 04Report like a humanCoverage data into a story.
- 05Learn GEOHow AI search reads PR.
How should a PR account team start learning AI?
1. Use AI for first drafts of releases and pitches
Give AI the facts, the angle, the agreed quotes and the client's style rules, and it'll hand you a workable first draft in under a minute. What it won't give you is news judgement. That's still yours, and it's the bit clients are paying for.
Build a voice sheet for every client: words they love, words they ban, how formal the quotes should be and how each spokesperson sounds. Paste it in every time. The difference between a beige draft and a usable one is almost entirely the brief.
For pitches, ask for five subject lines and three angles, then bin four of the subject lines. AI is generous with options and unreliable at choosing between them.
PR prompt to try
Using the facts and approved quote below, draft a 350-word press release for trade media. Follow the client voice sheet. Put the news in the first sentence. No superlatives and no 'excited to announce'. Then give me three alternative headlines, with one sentence on which publication each would suit.
Check your progress
2. Research and media lists: let AI find, then verify
AI is handy for background work: summarising a sector, spotting recent coverage themes and suggesting which kinds of outlet cover a subject. It's also perfectly capable of inventing a journalist, giving them a beat and a plausible email address, all with a straight face.
So treat every name, title and statistic as unconfirmed until you've checked it against the outlet, the byline or your media database. Tools with live web search and cited sources help. They don't remove the need to click the link.
An account executive who learns this in month one is worth far more than one who learns it after a client spots a made-up figure in a release.
Check your progress
3. Build a reactive comment workflow with proper sign-off
Reactive work is where AI saves the most stressful minutes. Feed it the story, your client's known positions and their past comments, and ask for three drafts at different temperatures: cautious, balanced and a bit punchy.
Then the human bit. The account lead picks, sharpens and sends to the spokesperson. Nothing goes to a journalist that the client hasn't approved, and nothing embargoed or market sensitive goes into a public AI tool. Write that rule on the wall if you have to.
PR prompt to try
Here is a breaking news story and our client's published position on the issue. Draft three 60-word reactive comments: one cautious, one balanced, one more opinionated. Flag any claim in each that the client would need to confirm before we send it.
Check your progress
4. Turn coverage data into a report somebody reads
Coverage reports are where good agencies go to sound like spreadsheets. AI can take your monitoring export, group coverage by message and outlet type and draft a plain-English summary of what landed, what didn't and what to try next month.
Check the counts yourself, because AI occasionally adds up like a raffle at closing time. Then add the insight only an account team has: why the trade press ignored the launch, or why one journalist keeps calling back. That's the paragraph a client forwards to their board.
Check your progress
5. Learn how AI search engines read your clients' coverage
Most PR training skips this part. ChatGPT, Gemini, Perplexity and Google's AI answers lean heavily on earned media when deciding which brands to mention. That makes good coverage more valuable than ever, and it makes PR teams the natural owners of generative engine optimisation, usually shortened to GEO.
Beginners don't need the whole discipline. Start by asking two or three AI tools the questions your client's customers would ask, then note which brands get named and which publications get cited. That one exercise tends to change how an account team thinks about media targets.
When you're ready to go further, our GEO for PR Agencies module and the 30-step GEO guide cover it properly, and how to write content AI will quote helps with the copy.
PR prompt to try
Act as a UK buyer looking for [client's category]. Recommend three providers and explain why. List the publications and sources you relied on.
Check your progress
What should each person on a PR account team learn first?
One session can cover the whole team, but each level needs different first skills and different warnings.
| Role | First AI skills to learn | Watch out for |
|---|---|---|
| Account executive | Drafting releases and social copy from a brief; research summaries | Unverified names and figures |
| Account manager | Pitch angles, reactive comment drafts, coverage report summaries | Skipping client sign-off under time pressure |
| Account director | Client voice sheets, quality review, AI disclosure conversations | Prompts drifting away from the brief |
| Agency head | Policy, approved tools, pricing AI-assisted work, GEO as a service | Selling AI efficiency the team hasn't learned yet |
Juniors gain speed first. Seniors gain the right questions to ask.
A first exercise for your account team
Pick one live client. The executive drafts a release from last month's news using the voice sheet. The manager writes three reactive comments on a current story. The director asks two AI tools who leads the client's category and records which outlets get cited.
Meet for thirty minutes, compare notes and agree three rules for using AI on that account. Repeat with a second client the following week and you'll have the bones of an agency policy written by the people who'll follow it.
Train your account teams with us
Our GEO for PR Agencies module shows account teams how AI search picks its sources, and Introduction to GEO is the gentler starting point. For the writing side, AI for Marketing Teams covers drafting, editing and campaign work. Harry's case for why PR should lead this is in PR agencies should be the GEO pathfinders.
Based in Witney, Oxfordshire. Explore UK training options or take the free two-minute AI skills assessment to see where your team stands.
Talk to us about your agencyAI training for PR agencies: frequently asked questions
Should PR agencies tell clients they use AI?
Yes, in plain terms. Explain which tasks AI helps with, what never goes into an AI tool and who checks the output. Clients increasingly ask agencies about AI use, so a clear answer is good business as well as good ethics.
Can journalists tell if a pitch was written with AI?
Often, because generic pitches sound alike. A pitch built on a specific angle for a specific journalist reads as human because a human did the important part. Use AI for the draft and your judgement for the angle.
Is it safe to put embargoed client news into ChatGPT?
Not into a free or personal account. Use an approved business tool with appropriate data settings, and keep embargoed or market-sensitive material out unless your agency policy and the client's terms clearly allow it.
What is GEO and why does it matter for PR?
Generative engine optimisation is the work of getting brands mentioned and cited in AI-generated answers. Those answers draw heavily on earned media, so the coverage PR teams secure now shapes what AI tools say about their clients.
How long does AI training take for a PR team?
A half-day session gets an account team using AI on real client work. Give it a month of practice on live accounts, then review voice sheets and prompts together. GEO skills take longer and suit a dedicated module.
Sources and further reading
Statistics, standards and programme links checked on 5 October 2026.