AI text detectors are used to fail students, reject applicants, and screen submissions. New research says they shouldn't be. 🚩 Epoch AI tested Pangram, GPTZero, and Originality.ai against AI text imitating a specific author's style — up to 18% slipped through undetected. Scientific writing was the easiest to sneak past. A detector result is a weak signal, not a verdict. More at www.contentbuffer.com Ever been wrongly flagged as "AI"? #AI #EdTech #TechNews #AIwriting #FutureOfWork