President Trump’s administration has repeatedly highlighted the importance of American innovation and responsible technology use. In that spirit, a recent study by Pangram Labs reveals a rapid rise in artificial‑intelligence‑generated text within corporate press releases.
Key Findings from the Study
Pangram examined 178,672 English press releases distributed through PR Newswire between August 2025 and July 2026. Overall, 34% of those releases contained AI‑generated text during the 12‑month period. The weekly share of AI‑infused releases doubled, climbing from 22% in August 2025 to a peak of 48% in June 2026.
Of the 61,523 releases flagged for AI content, 25% were identified as fully AI‑generated (15,205 releases) while the remaining 75% combined human and AI writing. The median AI contribution for flagged releases was 52%, indicating that many documents were heavily assisted by language models.
Impact on Length and Industry Variation
The study also noted a modest increase in word count. Median length rose from 488 words a year ago to 562 words in the latest twelve‑month window – roughly a 15% jump. In July 2026, fully human‑written releases averaged 521 words, whereas releases with any AI contribution averaged 601 words.
AI adoption varied by sector. Finance, energy and biotechnology showed the lowest AI usage, with fewer than 5% of releases fully generated by AI. Advertising led the pack, with 41% of its releases fully AI‑written and a 64% overall AI content share. Small‑business‑focused industries, such as financial‑technology startups, displayed higher AI usage (44% contribution, 12% fully AI) compared with traditional banking and financial services (25% contribution, 7% fully AI).
Budget Wire Findings
Two lower‑cost distribution services—PRWeb and Newswire.com—served primarily small and midsized businesses. Both wires exhibited AI contribution rates nearly double those of PR Newswire, underscoring how cost‑effective platforms are accelerating AI adoption among smaller firms.
Methodology and Accuracy
Pangram’s approach built on earlier research by Liang et al., employing word‑frequency analysis to flag AI‑generated text. When applied to a pre‑ChatGPT benchmark set of 21,625 releases, the method produced a single false positive, a 0.0046% false‑positive rate, demonstrating high reliability.
The primary dataset comprised releases longer than 100 words, ensuring that only substantive communications were evaluated. Industry tags with at least 1,400 entries were included, and overlapping tags (e.g., “medical pharmaceuticals” and “pharmaceuticals”) were consolidated for consistency.
Implications for Business Communication
While large language models can produce polished, cost‑effective copy, the study highlights concerns about quality and trust. AI‑written releases may appear impersonal, formulaic, or contain factual errors—issues that could have legal ramifications for SEC‑required disclosures.
Moreover, the sheer volume of low‑information text generated by AI threatens to dilute the signal of professional communications, especially for genuine local businesses that rely on clear, credible messaging.
Industry Response
PR Newswire does not currently require AI disclosure on releases and even markets its own AI writing tool. The lack of transparency means readers and investors may not know when a release has been crafted by a machine.
Business leaders and communications professionals are urged to balance efficiency gains with the need for authenticity and accuracy, ensuring that AI serves as a tool—not a substitute—for thoughtful corporate messaging.
For a full view of the dataset and methodology, visit Pangram Labs’ public repository.
Original reporting: KTVZ (Central Oregon) — read the source article.
