The AI PR Playbook: How to Use AI on Your First Draft Without Sounding Like a Robot
Category: AI in PR
Every journalist now has a word for AI-generated press releases. It is not a compliment. Here is how to be the exception.
Every journalist now has a word for AI-generated press releases. It is not a compliment. Generic lede, overloaded adjectives, CEO quote that could apply to any company in any industry. Journalists call it "AI slop." It gets filed in the trash folder before the first paragraph ends. If you're the person responsible for how your company looks to the press, or the person a client is trusting to represent them credibly, that outcome is not acceptable. Here's the thing: the releases that avoid this problem are also written with AI. The difference is the workflow, and specifically, knowing what AI should never do alone. What AI Gets Right in PR Writing Structure: AI is reliable at knowing what a press release should contain and in what order. The five-W lede, the executive quote, the supporting data paragraph, the boilerplate: AI handles this scaffold well. This alone saves 30 to 45 minutes on a first draft. Completeness checking: After you draft, AI is good at asking "did you include X?" Have you addressed the business impact? Is there a specific market context? Does the boilerplate contain accurate company description? This is the rote review work that humans miss when they're too close to the material. Variation generation: Need three different headline options? Two versions of the executive quote? Different angles for different publications? AI generates these quickly and gives you real choices instead of the one version you'd write if you were doing it alone. What AI Should Never Do Alone Specificity: AI doesn't know the specific detail that makes your story interesting. It doesn't know that your CEO left Goldman Sachs to build this. It doesn't know that you're the only company in this category that's SOC 2 Type II certified. It doesn't know the customer story that proves the claim. You have to supply all of this. Every piece of generic-sounding AI output is generic because the human didn't provide enough specific input. Voice: Executive quotes written by AI have a tell: they use full, formal sentences with consistent structure. Real executives use fragments, emphasis, unexpected word choices. The AI quote sounds like a press release. A real quote sounds like a person. If you use an AI-drafted quote, rewrite it in the executive's actual voice before it goes out. This is the step that determines whether your CEO or client trusts the output. Newsworthiness judgment: AI doesn't know what's actually interesting to journalists. It can write a release, but it can't tell you whether the story has legs. That judgment still belongs to a human who knows the beat, the journalist, and the news cycle. The Prompting Approach That Works Most AI-generated press releases are bad because the prompts are bad. "Write a press release about our Series A" produces generic content because that instruction is generic. Here's what works instead: Front-load the specifics: Before asking for anything, give the AI the raw material: founding story, specific metrics, customer name (with permission), what the money is for, the one thing you want journalists to remember. The output is only as specific as the input. Give structural constraints: "Write a press release with a lede that leads with the funding amount and the specific use case, a CEO quote that expresses a specific opinion (not excitement), and a closing paragraph with company boilerplate under 80 words." Constraints produce better output than open-ended requests. Edit for voice, not just accuracy: After the AI draft, your editing pass should focus on two things: does every claim have a specific fact behind it, and does the executive quote sound like a human being? Everything else is secondary. What a Marketing Leader or Agency Owner Should Never Let AI Do Alone If you're a marketing leader who doesn't want to hand a generic AI draft to your CEO, or an agency owner who can't afford to send a client something that reads like a template, here's the short list: Never let AI write the executive quote without a human rewrite Never let AI choose the lede without giving it the specific most newsworthy fact first Never let AI set the angle without telling it which journalist you're targeting and what that journalist covers Never skip the final read-aloud test: if it doesn't sound like a human being, it isn't ready The Always-Human-Edit Checklist Does the first sentence contain the most newsworthy fact? Is every adjective ("innovative," "revolutionary," "leading") replaced with a specific claim? Does the executive quote contain an opinion, not a restatement of the announcement? Is the company name spelled correctly everywhere? Does the boilerplate accurately describe the company as it exists today? AI-assisted PR writing is faster. It's also lower quality on average, because most people skip the human edit. The teams that win use AI for speed and humans for specificity. That combination produces releases that move fast and actually get read. 24HRPR's structured prompts are built to collect the specific inputs that make AI output credible, not generic. Try 24HRPR free and write your first release today.