For as long as we’ve worked in PR, a press release has had exactly one job: get a reporter to stop scrolling. Every line was built for that single reader. Tight, quotable, compelling media attachments, and front-loaded with the fact that mattered most. That instinct hasn’t gone away, it’s just aimed at a different audience now.
Most brands are still writing releases exclusively for journalists, at the exact moment AI answer engines have quietly become one of the biggest readers of that same document. Most PR teams haven’t adjusted a single habit for it.
The habit we learned in 2000s-era PR, and why it’s aging badly
Years ago, working inside a DC-based PR and communications shop, we learned to write a release the way a wire editor reads one: top down, ruthless, no fact sitting more than a sentence from the headline. That discipline still holds up. What’s changed is who else is reading with that same top-down logic.
AI answer engines parse a press release the same way an overworked wire editor does, hunting for the clearest, most self-contained statement of fact near the top, ideally with a source attached to it. A release written to charm a reporter into a follow-up call doesn’t serve that second reader. A release written to be the single clearest paragraph on the internet about a specific fact serves both readers at once, and that’s the actual shift.
That DC training also taught us the value of one sourced number over ten adjectives. It’s a habit worth reviving, because it happens to be exactly what AI systems reward, and it’s what most releases, thick with “thrilled to announce” and “game-changing,” skip entirely. It’s the same restraint we’ve argued for elsewhere (Silence sells): the fewer words a brand wastes, the more the ones it keeps actually land.
AI answer engines are already reading your releases, whether you planned for it or not
Google’s AI Overviews now reach an estimated 2 billion users a month, with ChatGPT close behind at roughly 800 million weekly (HubSpot State of Marketing Report, 2026). Traditional search volume is projected to fall by around a quarter this year as more of that traffic moves into AI-generated answers instead of a page of blue links. A press release that only exists to be read once by a human on deadline is optimizing for a shrinking share of its actual audience.
This isn’t a call to write for robots instead of people. It’s a recognition that the two readers want almost the same thing: the fact, stated plainly, with a real source behind it, before the color and the quotes. Journalists have always rewarded that. AI systems just made it non-negotiable, because a vague or unsourced release simply doesn’t get cited, pulled, or summarized; it gets skipped.
That’s not a reason to panic about search disappearing. It’s a reason to treat every piece of earned media, releases included, as something written to be understood and repeated by a system, not just skimmed by a person on deadline.
What writing for two readers actually looks like
The fix isn’t a format overhaul. It’s a short list of habits that most releases have let slide:
Lead with the fact, not the mood. “Company X announces record growth” tells a machine nothing checkable. “Company X grew revenue 34 percent year over year, driven by [specific driver]” gives both a journalist and an AI system something to quote directly.
Attach a real, checkable source to at least one claim in every release. This is the single easiest lever available to a small brand with no PR budget, and it’s the one most releases skip entirely.
Name a real person and let them say something with a point of view, not a platitude. Generic executive quotes (“we’re excited about this milestone”) get cut by editors and ignored by extraction systems alike. A specific, slightly opinionated line gets kept by both, and it’s the same instinct behind sounding credible on the record (Talk like you mean it).
Build in a short, plainly worded question-and-answer near the end for anything a reader would naturally ask next. It reads like a courtesy to a human skimmer and functions like a gift to an AI system looking for a clean answer to lift.
None of this requires new software or a bigger budget. It requires treating the release as a piece of structured, sourced writing again, instead of a formality that gets dashed off after the actual strategy work is done.
Here’s what that looks like in practice. A weak line reads:
“We’re thrilled to announce a partnership that will transform how our customers experience the category.”
A rewritten line reads:
“The partnership adds same-day fulfillment in 12 new markets starting in October, according to [named source].”
Same news, same length, but only one of them gives a reader, human or machine, anything to actually cite. That’s the whole exercise: run every sentence in a release through the question “could this be quoted on its own and still mean something specific,” and cut or rewrite the ones that can’t.
It’s also worth saying that no single release does this work alone. The same topical-authority logic that applies to blog posts applies here: a brand that sends one sharply sourced release a year won’t build much AI-citation history, but a brand that sends a handful a year, each one specific and sourced the same way, starts to look like a source worth trusting on the subject. Consistency is doing as much work as any individual line.
Questions worth asking before you rewrite your next release
Does this mean press releases matter more than they did a few years ago? In a sense, yes. As more research starts inside an AI answer engine rather than a search results page, a release that’s built to be extracted cleanly has a better shot at shaping how a brand gets described, whether or not a journalist ever picks it up.
Do we still need to pitch journalists directly, or is the release enough on its own? Direct pitching still matters, especially for anything that needs real judgment or context a machine can’t supply. The point isn’t to replace that relationship. It’s to stop writing the release itself as if it only has one possible reader.
What’s the fastest way to test this on an existing release? Pull up the last one your team sent. Find the most important fact. If it isn’t in the first two sentences, with a number and a source attached, that’s the fix to make first.
Where we’d start
None of this is complicated, which is exactly why it keeps getting skipped. The brands that get cited, quoted, and summarized by AI systems this year won’t be the ones with the biggest PR budgets. They’ll be the ones who treated a press release like a sourced, specific piece of writing again, instead of an afterthought with a boilerplate stapled to the bottom.
