Why it is so difficult to say whether a text was written by AI – even for AI

People and institutions are struggling with the consequences of AI-written texts. Teachers want to know whether students’ work reflects their own understanding; Consumers want to know whether an ad was written by a human or a machine.

Writing rules to govern the use of AI-generated content is relatively easy. Their enforcement depends on something much more difficult: reliably detecting whether a text was generated by artificial intelligence.

Some studies have examined whether humans can recognize AI-generated text. For example, it has been shown that people who frequently use AI writing tools themselves accurately recognize AI-written text. A panel of human reviewers can even outperform automated tools in a controlled environment. However, this expertise is not widespread and individual assessment can be inconsistent. Institutions that need consistency at scale are therefore turning to automated AI text detectors.

The problem of AI text recognition

The basic workflow behind AI…