Skip to content

Fact · Voice and AI

9percent

On OpenAI's challenge set, its AI text classifier incorrectly labeled 9 percent of human-written text as AI-written.

Checked against its source on October 17, 2026

In the source's own words

“while incorrectly labeling human-written text as AI-written 9% of the time”

New AI classifier for indicating AI-written text (OpenAI, January 31, 2023, with July 2023 update), Jan Hendrik Kirchner et al., OpenAI, 2023, Limitations section of the post.

Open the original
What it measures
This is the false-positive rate from the same challenge set as the 26 percent true-positive figure, so both numbers describe one test.
What it does not show
A false-positive rate of 9 percent is a figure from one test of an early classifier. It does not predict how often any current detector flags a real human essay, and it says nothing about dictated text.

Source

Cited on

OpenAI's classifier labeled 9% of human text as AI | Yap