How to Use AI as a PR Specialist, Not a Spokesperson

A PR specialist's week splits into two different kinds of work filed under one job title: producing material — press releases, pitch emails, talking points, media lists, boilerplate — and the part built on judgement and a relationship: deciding which journalist will actually care about a given story, and standing behind a public statement issued under an organisation's name. A general AI tool is genuinely fast at the first. It has no access to the second: it does not know a specific reporter's beat, has no history of accurate pitches to that person, and cannot read a live situation the way someone watching it unfold can. Will AI replace PR specialists? answers the wider question this post assumes: no, because none of the drafting speed below touches the relationship or the judgement call.

Where AI genuinely helps

• —A first-pass press release from a set of facts you already have, so you are editing rather than writing from a blank page.
• —A spread of pitch-email variants for different outlets from one core story, instead of rewriting the same pitch by hand for each one.
• —Building a media list from a beat description, then checking each name against recent coverage before you send anything.
• —A plain-language summary of today's coverage of a topic, so you spend review time on the judgement rather than the reading.
• —A first pass at boilerplate or talking points from decisions already made, for you to tighten into the final version.

Why the relationship does not move

A pitch landing has very little to do with how well it reads and a great deal to do with who sent it. A journalist who has been pitched accurate, relevant stories by the same person for years opens that person's email differently from a stranger's. Task-level research mapping real AI usage onto occupational data keeps finding the same shape across knowledge work generally: these tools help most with producing and formatting text against a known structure, and a press release has a known structure. Knowing which of fifty journalists to send it to, in what form, so it actually gets read rather than deleted, sits outside that.

A model can draft a clean pitch in a minute. It cannot tell you which of fifty reporters will actually read it, and it has no relationship with any of them to protect.

The failure that actually costs you

The failure mode that matters most in this job is a fabricated detail reaching a journalist under an organisation's name. Hallucination in large language models is well documented: a tool can produce a confident, specific-sounding statistic, quote or claim with nothing behind it, and it reads exactly like a verified one until someone checks. A press release is a public, attributable statement — once a reporter has it, a wrong number or an invented quote is not a draft to quietly fix, it is a correction to issue. Checking an AI answer when you are not the expert covers that discipline in general terms; in this job the check has to happen before anything reaches a journalist's inbox, not after it is already quoted back.

A worked example

Weak: "Write a press release announcing our new product."

Better: "Here are the five facts we can stand behind about this launch: [list them]. Draft a press release using only those facts, no superlative claims and no competitor comparison. Keep it under 400 words and flag any sentence where you had to infer something I did not give you." That version restricts the content to facts you supplied and asks the tool to flag its own gaps, rather than trusting an open question to stay inside the lines — the same instinct OpenAI and Anthropic both give in their own prompting guidance.

The confidentiality line

An unannounced deal, a draft crisis statement or embargoed launch details should not go into a general consumer tool with no agreement covering how it is retained. Deciding in advance what a draft may and may not see is the core idea behind the NIST AI Risk Management Framework — a line to set before a statement goes anywhere near a tool, not a cleanup after it already has.

Getting started this week

1. 01Get genuinely fast at the drafting layer — release first drafts, pitch variants, media list building from a beat description — because that is the real time saving and the lowest-risk part to hand over.
2. 02Build one fixed checkpoint before anything goes out: read every drafted release or pitch once for tone and once specifically for a number, quote or claim that was not in the original brief.
3. 03Write down, once, what never gets drafted by a tool at your organisation — a statement during an active crisis, anything responding to a direct, unexpected question from a reporter.
4. 04Invest the time a tool saves into the relationships a tool cannot build — the calls and follow-ups with the handful of journalists who actually cover your organisation's work.

The wider pattern this fits

This sits one level under how to use AI as a marketer and alongside how to use AI as a social media manager, which covers the identical drafting-plus-checking habit for a different public-facing format. How to use AI as a copywriter narrows the same split to the line-writing level, and what jobs are safe from AI covers why a trusted external relationship tends to be durable across professions generally.

Use AI to clear the real drafting grind — releases, pitch variants, media lists — fast. Do not let it decide which journalist to trust with a story, or let an unchecked claim reach a reporter under your organisation's name. PwC measured a 62% wage premium for workers with demonstrable AI skills, up from 57% the year before. Coursium teaches that checking habit directly: short lessons on your phone, a quiz that checks the point actually stuck, and a practice task in using general AI tools well. Stay ahead of AI rather than waiting to see which half of the job changes first.