The paper showed that asking chatGPT to reword an essay will evade the detector.
Assuming they know of the existence of GPT detectors, a student who uses AI to write or reword their essay is LESS likely to be flagged as a cheater than a student who never used AI at all.