Nearly half of all corporate press releases distributed through PR Newswire now contain AI-generated text, according to a new analysis from Pangram Labs. The findings show AI adoption in business communications has doubled over the past year, reaching 48% in June 2026. For PR and communications professionals, the shift signals a rapid rewiring of how companies approach the workhorse document of media relations.
The study scanned 178,672 English-language press releases published on PR Newswire between August 2025 and July 2026. PR Newswire is the largest press release distributor in the U.S., handling material from companies such as Pfizer, Toyota, and Walt Disney World, including mandatory SEC disclosures. Pangram found that 34% of releases over the 12-month period contained AI-generated text, with the weekly share climbing from 22% in August 2025 to 48% by June 2026.
AI authorship is substantial, not just light editing
When AI appeared in a press release, it was rarely a marginal assist. The median release flagged for AI content was 52% AI-authored. A quarter of all flagged documents - 15,205 releases - were fully AI-generated. The remaining 75% contained a mix of human and machine-written text. Across all flagged releases, the data showed a bimodal distribution: documents either used AI very lightly or relied on it almost entirely.
AI use also appears to be driving longer press releases. The median word count across PR Newswire increased about 15% over the twelve-month period, from 488 to 562 words. Fully human-authored releases in July 2026 had a median length of 521 words, while releases containing AI text averaged 601 words.
Industry splits show a tale of two sectors
AI adoption varied sharply by industry. Finance, energy, and biotechnology had the lowest rates, with fewer than 5% of releases flagged as fully AI-generated. Advertising sat at the opposite extreme: 64% of its press releases contained AI text, and 41% were fully AI-authored. Much of that volume came from content marketing agencies distributing editorial-style articles for local businesses. One such issuer, HelloNation, published 2,265 releases in the dataset. Of those, 2,198 were fully AI-generated - a 97% rate - and only one flagged as human-written.
The divide extended beyond industry labels. Sectors heavy with small businesses and startups showed far higher AI use than those dominated by large, established institutions. Financial technology releases, often issued by single-person startups and crypto ventures, had a 44% AI contribution rate. Banking and financial services, populated by traditional banks and funds, sat at 25%. Budget wire services that cater to smaller businesses showed similar patterns: both PRWeb and Newswire.com had AI contribution rates roughly double that of PR Newswire.
Previous studies missed the acceleration
An earlier study by Liang et al. used word frequency analysis to estimate AI prevalence in corporate communications. Pangram rescanned that same corpus and found the earlier method overstated early adoption. Instead of a rapid spike after ChatGPT's November 2022 launch, AI uptake was gradual. It took nearly a year for press releases to break a 5% AI contribution rate. The rate then hit 9% by August 2024, doubled to 22% by July 2025, and doubled again to 44% by July 2026. The older dataset also served as a benchmark for Pangram's detection tool: on releases predating ChatGPT, it produced one false positive among 21,625 documents, a 0.0046% false positive rate.
Why this matters for PR and communications professionals
None of the major wire services require AI disclosure. PR Newswire sells an AI writing tool and states it does not publicly tag releases as AI-generated. For communications teams, the quality risks are real. AI-written text can introduce factual errors - a serious liability for releases tied to SEC obligations. Research also shows audiences find AI writing less trustworthy, and fatigue with recognizable LLM phrasing continues to grow.
The volume problem may cut deeper. Large language models let anyone produce polished, grammatical text at near-zero marginal cost. The result, Pangram notes, is a world where "anyone can publish a press release, memo, or report essentially at the cost of distribution, and the length and frequency of these texts will increase accordingly." For PR professionals who need their announcements to cut through, the signal-to-noise ratio is deteriorating. Smaller players - genuine local businesses without massive distribution budgets - stand to lose the most as the channels fill with what Pangram calls "workslop." Professionals looking to navigate this shift can explore training on AI for PR & Communications and specialized resources for AI for Media Relations Specialists.
Your membership also unlocks:
