ChatGPT and Turnitin: The Rise of AI Detectors Fuels Distrust

ChatGPT and Turnitin: The Rise of AI Detectors Breeds Distrust

How It All Started

Long before ChatGPT became popular, educators and publishers frequently used anti-plagiarism tools to check the honesty of writers. These tools compare a written text to a database filled with content from the web, academic articles, and more, in order to spot similar phrases and sentences. Some, like Turnitin, provide a percentage that claims to illustrate the degree of overlap between a student's work and other works. Due to potential false positives and uncertainties regarding intentional duplication, some educators began to turn away from this tool.

However, the search for copied content is now evolving into a war against AI-generated works. Just as quickly as students adopted ChatGPT, Google Gemini, and Microsoft Copilot, teachers have integrated AI detectors. A survey by the Center for Democracy and Technology revealed that 43% of teachers in grades six through twelve in the United States regularly used AI detectors between 2024 and 2025. Some universities already using Turnitin in their learning management systems found that the service automatically activated AI detection upon its launch in 2023.

Instead of comparing pieces of written content, AI detectors like GPTZero, Pangram, and the one created by Turnitin rely on their own AI models to guess whether a text might not be written by a human — a process that is arguably even more opaque than text matching on the web. As GPTZero indicates, AI detectors use an algorithm to analyze word choice, rhythm, and structure of a text, as well as to spot patterns of length and tone that might be more common in AI-generated texts. This relatively subjective assessment is not as robust as something that could be justified by an online text match and can be misled by writers for whom English is a second language. Despite this, Turnitin claims that its AI detector falsely flags less than 1% of human-written content as AI, while Pangram asserts that its false positive rate is only 1 in 10,000. GPTZero also claims a similar rate of confusion between human and AI content.

How It’s Playing Out

Online users are already accusing each other of "sounding like an AI," but the availability of AI detection tools only adds fuel to the witch hunt against LLMs. In some highly publicized cases, accusations of AI writing have directly impacted individuals' livelihoods and reputations. Last month, the publisher Minotaur canceled a $2 million contract over concerns that its author, Jerry Falade, had used AI — a claim he vehemently denies.

There is also Thierry Rignol, a French citizen who sued Yale last year after a professor accused him of having written parts of his final exam with AI, resulting in a failing grade and a one-year suspension. The professor used GPTZero to analyze Rignol's writings for signs of AI, but the lawsuit argues that "AI monitoring and detection tools are known to unfairly target non-native English speakers" like Rignol. In February, a student from Adelphi University won a lawsuit against the institution after his professor also claimed he had used AI to write an essay. Although the lawsuit does not specify which AI tool the professor used to examine the student's essay, Adelphi University has a licensing agreement with Turnitin.

A Stanford study in 2023 found that AI detectors falsely flagged essays written by non-native English speakers as AI more often than those written by native speakers. (Many services still claim that their tools are accurate when it comes to texts written by non-natives.) These tools may also be biased against neurodivergent writers.

As highlighted by the University of California, Los Angeles, AI detection tools are trained to spot patterns that might indicate AI usage, such as repetitive terms and phrases, text that seems overly formal or informal, and nonsensical formulations. Some, like QuillBot, also measure the “unpredictability” of the text, as "AI tends to make the most 'obvious' or common language choices compared to writing produced by humans," according to UCLA. They may also look for a sentence structure that remains the same throughout the text as another sign of AI. But these measures are not indicative of AI alone, as some individuals may simply have a writing style with these characteristics.

Although Turnitin boasts low false positive rates, it maintains that its tool "may not always be accurate" and should not be used to take action against a student. Grammarly warns that users "should never rely solely on the results of an AI detector," while GPTZero claims "that no AI detector can ever be truly 100% perfect." OpenAI even shut down its own AI writing detector in 2023 due to its low accuracy.

However, accusations of AI writing continue to circulate online. Last week, in a video broadcast to over 3.5 million followers on his social media, Jack Osbourne, the son of Ozzy Osbourne, accused journalist and Verge contributor Kat Tenbarge of using AI to write an article for Rolling Stone, while showcasing the results of an AI detector, Getsolved, as "proof" of his claim. Tenbarge refuted this accusation in a video and a post on her site, but Osbourne did not retract his accusation or remove the video, leaving Tenbarge to face trolls.

There are numerous examples of false accusations, ranging from writers to students, with many accusers neglecting to acknowledge the warnings that accompany some of these AI detection tools.

What Happens Next

The uncertainty surrounding AI detectors is prompting some educational institutions to cease using them altogether. Yale, Johns Hopkins University, Vanderbilt University, Georgetown University, and others have disabled or restricted the use of AI detection tools. The Massachusetts Institute of Technology also warns that "AI detectors do not work."

Instead of relying on tools to detect AI, many schools are encouraging educators to rethink their lessons. The University of Chicago, for example, suggests asking students to slow down their reading, break long writing tasks into smaller parts, and require students to reflect on their work. Stanford University advises professors to consider the possibility of holding in-class assessments, while MIT recommends that professors allow students to disclose if they used AI to assist with a task, without penalty.

With the growing prevalence of AI in and out of the classroom, efforts to distinguish what is written by a human or a machine are also multiplying. Some online platforms exacerbate suspicions surrounding AI usage. Substack has integrated Pangram into its application, allowing users to scan blogs for content suspected of being AI-generated, while LinkedIn has added a "seems to be AI" button on posts. The result is a new era of distrust, where readers constantly question whether what they are reading is AI-generated and where human writers do their best not to sound like it.

About

The Authors Guild helps writers anticipate accusations by issuing "Human Authored" certifications. There are also "Not by AI" and "Written by Human" badges that users can add to their online works.

Wikipedia has created a guide to help editors spot AI writing, which includes elements to watch for, such as text that "inflates" the importance of a subject or provides a "superficial analysis of information." The site has also banned AI-generated articles.

The New York Times offers a fun quiz that asks you to compare five pairs of passages and choose which one you prefer — the catch being that one of them is written with AI.

Inside Higher Ed has interviewed some educators about their approach to AI in schools, with one speaker expressing concerns about the costs and resources needed to make assignments AI-proof. My colleague Jess Weatherbed has detailed how the fanfiction community is also going through its own internal struggle to determine which stories are AI-generated.