If a music writer opens ten press releases this month, they’ll spot something odd. They all sound like the same person wrote them. Not because the artists share a publicist, or because there’s a house style doing the rounds, but because a growing number of them were drafted, in whole or in part, by the same few AI prompts, all trained on the same diet of prior press releases, all juggling the same set of stock phrases.
My AUK colleagues will recognise the phrases. The record that “weaves together” influences. The songwriter “unafraid to bare their soul.” The album that’s “both a departure and a return to form.” None of these phrases are new; PR writing has always leaned into shorthand, but what’s changed is the speed and scale at which that shorthand is now being produced, and the effect that speed is having on the writing that surrounds music, from pitches for reviews to the artist bios on Spotify.
The pitch problem
A PR pitch used to be, and from the better publicists still is, a small act of persuasion. Someone who listened to the record, and then tried to explain why this particular songwriter, in this particular moment, deserved a few hundred words in a magazine that people trust. It required a point of view. Even a bad pitch had a person behind it you could disagree with.
What’s landing in our inboxes now increasingly reads like it was generated from a prompt containing the artist’s genre tags and a rough mood board. The record is “sonically rich.” The lyrics are “deeply personal.” The artist “draws on” a list of influences that are more about word association than the artist’s record collection. None of it is wrong, exactly. It’s just not saying anything meaningful. A pitch that could apply to fifty different records isn’t a pitch; it’s noise with a press date attached.
The irony of course is that AI tools are being used to save time on the part of the process that most benefited from taking time. Understanding a record well enough to describe it accurately, and interestingly, was never the bit you could or should shortcut.
What good PR does
It’s worth remembering what a strong pitch is for because usually the AI version misses the point of the job. A good pitch isn’t a summary. It’s an argument. Its task is to take a record that a reviewer, a programmer, a playlist curator or a radio producer hasn’t heard yet and give that person a reason to feel something about it before they’ve pressed play. Not a list of genre comparisons. A feeling.
The best pitches I’ve read do this by finding the one true, specific, slightly strange detail that unlocks the whole record: the songwriter wrote the album in the six weeks after their father’s death and you can hear it in how the drums drop out; the whole thing was recorded live in a barn in an afternoon because the band couldn’t afford a second session; the title track is about a bar that’s since been demolished. That detail does the persuasive work that adjectives can’t. It gives the person on the receiving end something to hold onto, and a way to describe the record to their own readers or listeners that isn’t generic.
This is the actual craft of PR writing: translating an emotional experience into language that reproduces some of that emotion in a stranger who’s never heard the music and has forty other emails to get through before lunch. It’s closer to writing a short story than to writing a product description. It has to be persuasive in the way a friend’s recommendation is persuasive, because it comes from somewhere real.
Why this hits niche genres especially hard.
Americana, or indeed Metal or Folk or a dozen other genres, sell themselves on specificity. Nobody comes to our corner of the musical forest for polish; they come for the grain in someone’s voice, a story of life in a particular town or family, production choices that sound handmade rather than optimised. The whole appeal of Americana is that it resists being smoothed out.
Why AI dampens the wrong thing.
This is where AI writing on a PR pitch does the most damage, and it’s not really about grammar or fluency. It’s about what the technology is optimised for. A large language model is built to predict the next most likely word based on everything it’s seen before. That’s a good approach for a weather report. It’s a terrible way to persuade an editor to review your work. That’s because persuasion depends on surprise, specificity and a point of view that couldn’t have been predicted from the average of everything written before it.
Emotion in writing tends to come from precision, not sentiment. It’s the detail that couldn’t apply to any other record. AI text reaches instead for the emotional trigger word without the specific cause, so you get “raw,” “haunting,” “deeply moving,” attached to nothing in particular. Strip out the concrete detail and the feeling goes with it. The reader senses they’re meant to feel something, without ever being given a reason to.
A dull press release is a wasted afternoon. But a press release that fails to move the person deciding whether a record gets reviewed, played or playlisted is a record that never gets its shot, no matter how good it actually is. The technology isn’t just making the writing worse. It’s quietly making it less effective at the one thing it exists to do.
To AI or not to AI
None of this means AI has no place near a keyboard. Used properly, it’s a decent first-draft machine for the parts of the job that were always mechanical: chasing down release dates, checking tour listings, tidying a transcript. The trouble starts when it’s used as a substitute for the part of the job that was never mechanical, which is having something to say and the nerve to say it plainly.
The writers and publicists who’ll still matter in five years are the ones whose copy you could pick out of a line-up. At AUK we are lucky that most of the specialist promoters are exceptionally good at their job and super keen to get their artist in front of us. It’s when a non-specialist or maybe the artist themselves “has a go” at a bit of PR that it all falls apart.
AI technology isn’t going anywhere; there’s no putting it back in the box. But readers, and editors, are getting quicker at spotting the difference between a piece that was considered, crossed out and rewritten and a piece someone generated with a couple of button pushes.
Great article Tim