Technology
Pew Research finds growing share of web pages contain AI-generated text
An analysis of English-language websites indicates rising AI authorship since late 2022, with commercial domains showing the highest concentration.
The short version
- A Pew Research study analyzed roughly 500,000 English-language web pages and found that over one-third of pages published after ChatGPT's release show signs of AI authorship or editing.
- Across the broader five-year web sample, about 10% of pages exhibited significant markers of AI generation.
- AI-generated content is unevenly distributed, appearing on about 10% of .com pages compared to around 1% of .edu and .gov domains.
- Researchers cautioned that automated AI detection tools carry inherent inaccuracies, including false positives and false negatives.
Key facts
- Pew researchers analyzed approximately 500,000 English-language pages from Common Crawl over a five-year period using the Open Pangram AI detection tool.[CNET · Slashdot]
- In a July 2026 web snapshot, over one-third of pages published following ChatGPT's release showed evidence of AI generation or editing.[Slashdot]
- Across the entire five-year dataset, approximately 10% of sampled pages displayed significant indicators of AI authorship.[CNET · Slashdot]
- In 2026 samples, AI-derived text appeared on roughly 10% of .com domains, compared to 4.6% on .org domains and about 1% on .edu and .gov domains.[CNET · Slashdot]
- Stylistic changes identified in online text since 2023 include a doubling of em dashes, a 63% increase in Oxford commas, and a near-tripling of negative parallelism phrasing.[CNET · Slashdot]
What remains uncertain
- The exact proportion of AI-written content is subject to error because detection software can misidentify human writing as AI-authored and fail to detect synthetic text.[CNET]