AI Detectors Are Failing to Spot Mimicked Writing

Experts recently tested three popular AI detection tools. They wanted to see if the software could catch text written by AI that was instructed to copy a human writing style. The results show the detectors are not as reliable as many people think.

In some cases, the detectors failed to identify AI writing nearly twenty percent of the time. When it came to scientific papers, the failure rate jumped to almost fifty percent. This is concerning because schools and journals rely on these tools to verify academic work.

These findings suggest that you cannot fully trust AI detection software today. As models get better at mimicking human tones, these tools are becoming less effective. It is clear that this technology still has a long way to go.