AI Detectors Fuel Institutional Distrust

Educators and publishers are increasingly deploying AI detection tools to police written work, contributing to what The Verge describes as an "LLM witch hunt." As first reported by The Verge, a Center for Democracy and Technology survey found that 43 percent of U.S. teachers in grades six through 12 regularly utilized these detectors between 2024 and 2025.

Unlike traditional plagiarism tools that match text against databases, detectors from companies such as GPTZero, Pangram, and Turnitin use AI models to analyze rhythm, structure, and tone to guess if a work is human-written. While Turnitin claims a false positive rate of less than 1 percent and Pangram claims a rate of 1 in 10,000, these tools have led to severe real-world consequences. The Verge reports that the publisher Minotaur canceled a $2 million book deal with author Jerry Falade over AI concerns, which Falade denies. Additionally, Thierry Rignol sued Yale after a professor used GPTZero to accuse him of using AI on a final exam, leading to a suspension.

Critics argue these tools unfairly target specific groups. Research from Stanford in 2023 indicated that AI detectors misidentify essays from non-native English speakers as AI-generated more often than those from native speakers. The Verge further notes that these tools may be biased against neurodivergent writers. Despite these risks, some universities found that Turnitin automatically enabled AI detection features upon the tool's 2023 launch.