Pew study: a third of new web pages show signs of AI writing
490,000 pages analyzed: since ChatGPT's launch, over a third of new pages show machine-written traces — .com sites ten times more than .edu.
Illustration · AI-generated (AI IN LIFE)
At a glance
- Study: Pew Research Center, published 20 August 2026, widely covered on 24 August
- Data: 490,000 English-language pages from Common Crawl (Jan 2021–Jul 2026), tool: Open Pangram
- Result: ~10% of all pages show AI traces; over one third of pages since ChatGPT's launch (Nov 2022)
- Domains: .com ~10%, .org 4.6%, .edu/.gov ~1% each
- Language shift since 2023: em dashes doubled, Oxford comma +63%, AI vocabulary more than doubled
How much of the web does the machine now write? Pew Research Center delivers the first robust numbers: about 10 percent of all English-language web pages examined show significant signs of AI authorship — and among pages published after ChatGPT's launch in November 2022, it is more than one third. The basis is an analysis of 490,000 pages from the Common Crawl archive (January 2021 to July 2026) using the Open Pangram detection model, published by Pew Research.
The distribution is telling: commercial .com domains sit at around 10 percent AI share, .org at 4.6 percent — while .edu and .gov pages are at only about 1 percent each. AI text is above all a phenomenon of the commercial web, roughly ten times more common than on educational and government sites.
Language itself is shifting measurably. Since 2023, em dash usage has doubled, the Oxford comma is up 63 percent, typical "AI vocabulary" such as delve, pivotal and meticulous appears more than twice as often, and constructions like "it's not just X, it's Y" nearly three times as often.
One caveat matters, as The Decoder highlights: "Current detection tools can barely tell the difference between fully automated text and writing that was only partly AI-assisted." The one-third figure measures traces, not full automation.
For publishers, SEO and anyone seeking visibility online, the consequence is clear: AI text is the new normal, and provenance and labeling become the differentiator. AI in Life openly labels its AI-assisted production for exactly this reason.
FAQ
How was it measured?
Pew analyzed 490,000 English-language web pages from the Common Crawl archive using the Open Pangram machine-learning detector.
Does this mean a third of the web is fully AI-written?
No. The study measures signs of AI involvement. Tools can barely distinguish fully automated text from AI-polished human writing.
Where is AI text most common?
On commercial .com sites (about 10 percent) — roughly ten times more often than on .edu or .gov sites (about 1 percent each).
