TechCrunch: one in three post-ChatGPT web pages shows signs of AI authorship
Pew Research Center and the AI detection company Open Pangram analyzed nearly 500,000 English-language web pages published after November 2022 — when ChatGPT launched — and found that approximately 35% carry signs of AI authorship. TechCrunch covered the study on August 20, 2026.
The methodology used Open Pangram’s detection technology on the Common Crawl web archive. The distribution varies significantly by domain type: .com domains showed approximately ten times higher AI authorship rates than .edu or .gov sites, while .org domains measured closer to 4.6%. The researchers acknowledged that detection tools can misclassify individual pages, but argued the aggregate data is directionally accurate at scale across 500,000 pages.
The study measures signal, not quality. A page classified as AI-authored may be well-written and factually reliable; one classified as human-authored may not be. What the data documents is the structural change in the origin of newly published web content since late 2022 — a shift that has happened largely without formal acknowledgment from either publishers or platforms.
For content teams, the implications are practical. First, the volume of AI-generated material competing for the same search positions and AI-assistant citations is substantial and growing. Human-written content does not automatically rank higher: search algorithms and AI answer engines evaluate quality, authority, and freshness, not authorship method. But the growing baseline of AI-generated content changes what “standing out” requires — generic and formulaic material faces more direct competition, while content with a distinct point of view, primary research, or genuine expert perspective holds more relative value.
Second, the study provides context for editorial policy discussions. Content leaders debating internal AI use policies now have a data point about the broader landscape. If 35% of the web since ChatGPT is AI-authored, editorial decisions about disclosure and AI use are being made inside an environment where those norms, quality signals, and competitive dynamics are all shifting simultaneously.
A separate Graphite study from earlier in 2026 found that AI-authored articles had reached near-parity with human-authored articles in newly published content, suggesting the share continues to grow beyond the 35% figure for the cumulative post-ChatGPT period.
Useful for content strategy leads, editorial directors, and anyone setting AI use policies inside a publishing or content-production organization.