AI detectors are reshaping how educators, editors and readers assess written work, ushering in a period of heightened suspicion over who, or what, produced a piece of text. Long before ChatGPT arrived, teachers and editors relied on anti-plagiarism tools to verify that writers were being honest. These systems compare a piece of writing against a database of web content, scholarly articles and other material to flag matching sentences and phrases. Some, such as Turnitin, return a percentage intended to show how much of a student’s work overlaps with existing texts. Because of the risk of false positives and uncertainty over whether duplication was intentional, some educators have stepped back from using that particular tool.
The search for copied content has now evolved into a wider effort to identify AI-generated work. As students have adopted ChatGPT, Google Gemini and Microsoft Copilot, teachers have taken up so-called AI detectors at a similar pace. A survey from the Center for Democracy and Technology found that 43 percent of sixth to 12th grade teachers in the US regularly used AI detectors between 2024 and 2025. Some universities already running Turnitin within their learning management systems found the service automatically enabled AI detection when the feature launched in 2023.
How the detection works
Rather than comparing pieces of written content, tools such as GPTZero, Pangram and Turnitin’s own detector rely on their own AI models to estimate whether something may not be human-written, a process arguably murkier than matching text on the web. GPTZero has noted that these systems use an algorithm to analyse a text’s wording, rhythm and structure, and to pick up on patterns in length and tone that may be more common in AI-written material. This relatively subjective evaluation is less solid than results backed by matching text online, and it can be tripped up by writers who speak English as a second language.
Despite these limitations, Turnitin has said its AI detector falsely flags less than 1 percent of human-written content as AI, while Pangram claims a false positive rate of just 1 in 10,000. GPTZero says its rate of mistaking human content for AI is similarly low.
Real-world consequences
People online are already accusing one another of “sounding like AI”, and the ready availability of detection tools is adding fuel to the scrutiny. In some high-profile cases, AI writing accusations have directly affected livelihoods and reputations. The publisher Minotaur dropped a book deal worth roughly £1.5 million over concerns that its author, Jerry Falade, had used AI, something he strongly denies.
Thierry Rignol, a French national, sued Yale after a professor accused him of writing portions of his final exam with AI, resulting in a failing grade and a one-year suspension. The professor used GPTZero to scan Rignol’s writing, but the lawsuit argues that AI detection tools are known to unfairly target non-native English speakers. In February, a student at Adelphi University won a lawsuit connected to such an accusation.
Source
Image: theverge.com
