How Journalists Use AI to Screen PR Pitches

Most pitches now get filtered by a machine before a human ever reads them. Reporters use AI to summarize inbound email, flag generic or bot-written pitches, and fact-check claims in seconds, which means the pitch you send has to survive an algorithm’s judgment before it earns a journalist’s attention.

Why this shift happened

Inbox volume is the driver. Reporters are getting more pitches than they can physically read, and AI tools are the only way to keep up. According to Muck Rack’s 2026 State of Journalism report, 82% of journalists now use at least one AI tool in their work, up from 77% the year before. ChatGPT leads at 47% adoption, Gemini has climbed to 22%. The same report found that 86% of journalists say PR pitches inspire at least some of their stories, but 88% delete pitches that miss their beat. AI is what lets them make that call faster.

Most of that AI use is not for writing. It is for triage: researching a source, summarizing a long pitch into three lines, checking whether a stat holds up. That is the exact layer your pitch now has to pass through.

What AI screening actually catches

Reporters are not running your pitch through a formal “reject” algorithm. They are using AI the way anyone uses it to cut through noise, and it is very good at spotting a few specific things.

Generic, templated language. A pitch that reads like it was built from a mail-merge template gets flagged fast, either by the reporter’s own instinct or by AI summarization tools that strip a pitch down to its core claim and expose how little is actually there. The expert-sourcing platform Qwoted now runs pitches through the Pangram AI detector so reporters can see how much of a pitch was machine-written, and Qwoted’s analysis of more than 54,000 pitches reportedly found an average AI-likelihood score of 33%, according to journalist Joni Sweet.

Off-beat targeting. AI makes it trivial for a reporter to check whether a pitch actually matches their recent coverage. If it does not, it is gone before a human even weighs in.

Unsupported or shaky claims. Reporters increasingly use AI to fact-check a stat or a claim in a pitch before deciding whether to engage. A pitch that cites a number without a source, or that describes a “first” or “only” that is not actually true, gets caught quickly and quietly ignored.

Volume itself. AI writing tools make it cheap to send more pitches, and reporters see the results in their inboxes. That volume has made reporters faster and less forgiving screeners across the board, not just for AI-written pitches.

What this means for how you pitch

The practical upside here is that AI screening rewards exactly the things good pitches always needed: specificity, real data, and a story a human editor would actually run. It just enforces those standards faster and more consistently than a tired reporter scanning an inbox at 7am used to.

A few things matter more now than they did three years ago.

• Lead with the specific, not the category. “We help businesses grow” fails both the human read and the AI summary. A number, a name, a date, or a result in the first sentence survives both.
• Cite a real source for every claim. If you reference data, link to where it came from. Reporters are fact-checking claims with AI in seconds, and an unsourced stat now reads as a red flag rather than a shortcut.
• Match the beat before you hit send. Do the five minutes of research. A pitch that is clearly built for that reporter’s actual coverage reads as different from the templated batch the moment an AI tool summarizes it.
• Cut the boilerplate company description. Anything that sounds like it was copied from a press release template gets flagged as generic almost instantly. Say what is actually new.
• Write like a person, not a pitch generator. Ironically, the more AI-polished and generic a pitch sounds, the more likely it is to get caught by the same tools built to catch AI content. A little roughness and a clear point of view now read as credibility signals.

For subject lines and full email templates built around these rules, see our guide on how to pitch a journalist.

None of this is really new advice. It is the same discipline that has always separated a pitch that gets covered from one that gets deleted. What is new is that the filtering happens faster and with less patience, because the reporter is not reading every line themselves anymore.

Where this fits into a bigger PR strategy

Surviving pitch screening is a tactic, not a strategy. If your outreach depends on cold pitches with no prior relationship or track record, AI screening will only make that harder over time as volume keeps climbing and reporters keep tightening their filters. The founders and companies who consistently land coverage are usually the ones who treat media relations as an ongoing relationship-building function, not a one-off blast when they have news.

If you are early-stage and building a media presence from scratch, our founder-led PR playbook walks through how to build that foundation before you need it. And if you want a sense of how we approach this for early-stage teams specifically, see how we work for founders.

The takeaway

Write every pitch as if a machine will summarize it into two sentences before a human reads it, because increasingly, one will. Lead with your most specific, most verifiable claim, cite your sources, and skip the boilerplate. If your pitch cannot survive being compressed to its core fact, it will not survive a reporter’s inbox either. If you want help building pitches and a media strategy that clear both bars, get in touch.

Frequently asked questions

Can journalists tell if a PR pitch was written by AI?

Often, yes. Generic, polished, templated language is easy for reporters to spot, and some platforms now check for it directly: Qwoted runs pitches through the Pangram AI detector so reporters can see how much of a pitch was machine-written. Even without a detector, AI summarization strips a pitch down to its core claim, which exposes pitches that have little specific to say.

Is it okay to use ChatGPT to write a PR pitch?

Using ChatGPT for research or a rough first draft is fine, but the pitch you send should be rewritten by a person with specific facts. Unedited AI drafts tend to sound generic, and generic pitches are exactly what reporters and screening tools filter out. Replace broad claims with a number, a name, a date, or a result, cite the source for every stat, and add a clear point of view.

Why do journalists ignore PR pitches?

Journalists mostly ignore pitches that miss their beat, sound templated, or make claims they cannot verify quickly. With AI tools, a reporter can check a pitch against their recent coverage and fact-check a statistic in seconds, so weak pitches get dropped before anyone reads them closely. Pitches that name a specific, sourced finding in the first sentence and clearly fit the reporter’s coverage are the ones that get opened and answered.

How long should a PR pitch be?

A PR pitch should be short enough that its main point survives being summarized into two sentences. In practice that means a specific subject line, a first sentence with your most verifiable claim, a few lines of supporting detail with sources, and no boilerplate company description. If the reporter wants more, they will ask. Long pitches mostly give an AI summary more filler to cut and a busy reporter more reasons to stop reading.