AI detection tools are creating a new era of distrust in classrooms and online writing spaces. On Aug. 9, U.S. time, the Verge reported that teachers or editors once used plagiarism checkers such as Turnitin to see whether work was copied. Those tools compare sentences with web documents and academic materials and show how much overlaps. But as generative AI such as ChatGPT, Gemini and Copilot spreads, monitoring is shifting to attempts to identify AI-written text.

A survey by the Center for Democracy & Technology found that 43 percent of U.S. teachers in grades 6 to 12 regularly used AI detectors in 2024 to 2025. Tools such as GPTZero, Pangram and Turnitin's AI detection function estimate whether text was written by a human or by AI by analyzing phrasing, rhythm, structure, length and tone. The problem is that this judgment is far less transparent than plagiarism checks that directly match sentences on the web. Concerns are being raised that work by students who use English as a second language or by neurodiverse writers could be mistaken as AI-generated.

Cases of harm are also emerging. Publisher Minotaur cancelled a $2 million publishing contract after suspecting that author Jerry Falade (제리 팔라데) used AI, but the author strongly denied it. Thierry Rignol (티에리 리뇰), a French national, filed a lawsuit after a Yale professor used GPTZero to judge that part of a final exam was written by AI and imposed a failing grade and a 1-year suspension. An Adelphi University student also won a lawsuit against the school over similar suspicions of AI use.

A Stanford study in 2023 said AI detectors more often wrongly classify essays by non-native English writers as AI-generated than those by native speakers. UCLA also explained that repeated expressions, overly formal or informal styles and monotonous sentence structures can be treated as AI traits, but those characteristics do not necessarily mean AI was used. Some natural writing styles can overlap with the standards used by detection tools.

AI detection companies also acknowledge limits. Turnitin has said its tool may not always be accurate and should not be used as the sole basis for student discipline. Grammarly and GPTZero also warned that users should not rely only on detection results. OpenAI also halted its own AI writing detector in 2023 because of low accuracy.

Because of such uncertainty, some universities including Yale, Johns Hopkins, Vanderbilt and Georgetown have disabled or restricted the use of AI detection tools. MIT has also stated that AI detectors are not effective. Instead, schools are changing course design by breaking long writing assignments into stages, requiring students to explain their process and increasing in-class assessments. Some are also calling for an approach that treats AI use as an issue of disclosure and reflection rather than punishment.

Online platforms are also deepening suspicion. Substack integrated Pangram into its app, and LinkedIn introduced a reaction button reading, "Looks like AI slop." As a result, readers are left questioning whether each text was written by a human, and human writers are facing a paradox in which they try not to look like AI.