AI detectors yield mixed results in identifying human writing
The Authors Guild tests five AI detectors on human-written texts with varied results.

The Authors Guild tested five AI detectors on human-written texts. Pangram and Grammarly correctly identified all of them, while Sidekicker and ZeroGPT flagged human-written articles as AI-generated. But the Guild also warns of a paradox: professionally written texts look statistically similar to AI output because language models were trained on exactly that kind of writing.
The results highlight inconsistencies in AI detection technology. The tests suggest that some detectors are more effective than others in distinguishing between human and AI-generated content. The paradox identified by the Guild underscores a challenge in AI detection.
Since language models are trained on professionally written texts, these models can produce output that resembles human writing. This similarity can make it difficult for detectors to accurately identify AI-generated content. The findings have implications for the use of AI detectors in various applications, including content moderation and authenticity verification.
The Guild's test results indicate that the choice of detector can significantly impact the accuracy of AI-generated content detection. The development raises questions about the reliability and consistency of AI detectors. As AI-generated content becomes more prevalent, the need for accurate detection technology will continue to grow.
Why this matters: The mixed results from the Authors Guild's tests have significant implications for the broader industry. For developers, the findings highlight the need for more effective and reliable AI detection technology. Businesses and consumers also stand to be affected, as the accuracy of AI detectors can impact the authenticity and trustworthiness of online content.
As AI-generated content becomes increasingly sophisticated, the challenge of distinguishing between human and AI-generated content will only continue to grow. The paradox identified by the Guild also raises questions about the potential for AI detectors to be gamed or manipulated, and the need for ongoing evaluation and improvement of these technologies.
Source: The Decoder