Testing of detection tools for AI-generated text

In short

Can current tools tell human writing from chatbot text? On their own document set, the authors judge the tools they tried inaccurate and unreliable, inclined to label text human, and weaker after obfuscation.

What it looked at

An evaluation of 12 publicly available tools and two commercial systems used in academic settings. The authors tested accuracy and error types on human text and chatbot text, and whether machine translation or obfuscation changed the results.

What it means for teaching

A caution against treating a tool's output as proof about any one student, based on the authors' own document set.

Limits

Results apply to the tools, versions and original document set the authors tested, and how the texts relate to real student work is not described.

Cited 926 times by other papers, as counted by OpenAlex in November 2026. A count measures attention, not quality.