Unverified60% confidenceBenchmarkExact time
OpenAI's AI text classifier has a true positive rate of approximately 26% and a false positive rate of approximately 9% for AI-generated text.
2
Sources
60%
Confidence
Long-term
Relevance
8/13/2026
First Seen
Sources
Related Claims
VerifiedOpenAI launched an AI text detector in 2023, but was forced to take it offline because its accuracy was only approximately 26% with a false positive rate of approximately 9%.78% similarUnverifiedFreeHumanizerAI's public ZeroGPT tests show that an AI-written article dropped from 100% AI to 0% AI after Ultra mode processing, a business email scored 0%, and a product description still scored 70%, with results depending on the text content.71% similarUnverifiedThe accuracy of text-feature-based AI detection tools (such as GPTZero and ZeroGPT) faces serious challenges from the latest generation of AI-generated content.70% similarUnverifiedOpenAI's detection method works by comparing whether text conforms to the probability patterns embedded during model generation, and outputs the likelihood that the text was AI-generated rather than a definitive conclusion.69% similarUnverifiedThe characteristics for identifying low-quality or misleading AI-generated content include: templated phrasing, links with promotional parameters, and avoidance of specific data.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/742478API
curl https://kongchang.com/api/v1/knowledge/claims/742478MCP
get_claim(id=742478)