AI writing hits a ceiling

"The plateau indicates that the feared takeover of human online writing by AI hasn't materialized - at least not yet. Why it matters: Researchers who've studied the spread of AI-written articles warn that once models start training on that content, the internet could become a massive feedback loop of low-quality, machine-generated content."

""These models are smart because of all the information we put on the web that was created without these models," Dan Klein, a UC Berkeley professor and AI model CTO tells Axios. "If we stop creating knowledge that is independent of these models, what's going to fuel that?""

"Within a year of ChatGPT's release, primarily AI-generated articles made up 35.9% of new online articles. Within two years, they reached 48%. But since early 2025, the share has hovered at around half of new articles."

"Reality check: Counting AI-generated writing is still messy. Many articles are no longer written purely by humans or AI. A human may use AI for outlining, drafting, rewriting or editing, making the line between "AI" and "human" fuzzier than the chart suggests. Graphite classifies an article as primarily AI-generated only when most of its text is detected as AI-written or AI-assisted."

AI-generated writing has not yet taken over online publishing, with its share rising after ChatGPT’s launch and then leveling off. Researchers warn that if AI models train on AI-written content, the web could become a feedback loop producing low-quality machine-generated material. Graphite sampled 55,400 English-language URLs from Common Crawl, filtered for article-like pages with sufficient length and publish dates from 2020 to 2026, and classified them using AI-detection tools. AI-generated articles rose to 35.9% within one year and 48% within two years, then hovered around half since early 2025. Classification remains uncertain because many pieces mix human and AI assistance, and detection tools can be imperfect.

Read at Axios