When we shipped version two of our project-management tool, our entire marketing budget for the launch was $2,000 and our list of media contacts was a blinking cursor in an empty spreadsheet. Every guide I read online said the exact same two things: hire a PR agency (the retainers I was actually quoted started at $12k a month, which is six times our whole budget), or spend six months patiently befriending journalists before you have anything to announce. We were launching in two weeks. Neither door was open. So I did the slightly unreasonable thing — I wrote our own press release with the Press Release Writer — A Real News Angle From a Product Announcement prompt, and then spent a fortnight cold-pitching it like my rent depended on it, which it sort of did.
I want to be honest that I went in skeptical. I'd seen AI-written press releases before — they read like a thesaurus had a panic attack, all 'thrilled to announce' and 'revolutionary new paradigm.' I half expected to spend more time fixing the output than I'd have spent writing it from scratch. That's not what happened, mostly because of one design decision in the prompt.
The angle was the whole game, and the prompt forces it first
Before it writes a single sentence of the release, the prompt makes you name the news angle. That one forced step saved me from the most common and most fatal founder mistake: believing your feature is the news. It is not. Nobody outside your company is excited that your software now does a thing. I gave GPT-4o our announcement details — a new AI-powered scheduling feature, a setup time we'd cut by about 40% in internal testing, and a co-marketing partnership with a mid-sized project-management consultancy. It correctly, immediately identified the partnership as the angle, not the feature. A named third party with its own reputation makes a release newsworthy; a feature on its own makes it a changelog entry that journalists have a folder rule to auto-archive. The lede GPT-4o produced was tighter and more newsworthy than anything I'd have written at 11pm running on launch-week adrenaline.
The flag that saved me from myself
The prompt slapped a [VERIFY BEFORE SENDING] flag directly onto the '40% faster setup' claim, and I am genuinely grateful it did, because I had that number from an internal test with twelve users. Twelve users is a promising early signal. It is absolutely not a publishable, defensible, will-survive-a-journalist's-follow-up statistic. I softened the claim to 'up to 40% faster in our own testing,' which is honest, specific about its source, and still perfectly strong. If I had wired the bare '40% faster' into the release, the first sharp reporter to reply would have asked 'faster than what, measured how, across how many accounts?' and I'd have had nothing but a guilty pause. The flag turned a potential credibility landmine into a footnote.
The pitch, by the numbers
Writing the release was half the job. The other half was the cold pitch, and the release made that half easier because I had something professional to attach instead of a rambling email.
- Sent the release to 28 journalists across the project-management, SaaS, and productivity beats
- Hand-wrote the first two lines of every single email referencing that reporter's recent coverage — zero mass-blast, because a blast reads like a blast
- Six replied within 48 hours, which for cold media outreach is a frankly shocking hit rate
- Three of them ran some version of the story within the launch window
- One placement in a trade publication drove 340 trial signups in a single week, our best week of the quarter
Why a robot-written release actually worked
Here's the mechanism, because it's worth understanding rather than just celebrating. Journalists know the press-release format intimately because it is the format — they've read ten thousand of them. They scan the lede first, the quote second, and bail instantly if either is weak. The prompt produces a release that respects that exact scanning behavior because it's structured the way editors unconsciously expect, with the news up top and the adjectives nowhere. I still did real work: I rewrote the CEO quote to sound like how she actually talks rather than like a press kit, and I personally verified every claim carrying a flag. But the bones — the structure, the lede, the angle — were right out of the box, and the bones are the part that takes experience to get right.
A quick dose of honesty so you don't think this is magic: a handful of outlets never replied at all, and one of the three that ran the story cheerfully mangled our product name in the headline, which is a hazard of going agency-free that no prompt can fix. Still — six replies, three placements, zero retainer, two weeks of work. I will take that trade every launch. Grab the Press Release Writer — A Real News Angle From a Product Announcement prompt on Prompt Dock and run it on your own next announcement, then pair it with my Webinar Promo Copy Pack — Email, Social, and Reg Page in One Pass prompt if your launch needs an event to hang the whole story on.