AI Detectors Spark Growing Distrust in Gaming Community

This is The Stepback, a weekly newsletter diving into a key story from the tech landscape. For insights on how AI is reshaping our everyday experiences, keep an eye on the latest updates. The Stepback lands in subscribers’ inboxes at 8 AM ET.

How it started
Before ChatGPT took the spotlight, educators and editors often turned to anti-plagiarism tools to verify the authenticity of written work. These tools scan a vast database of online content, academic papers, and more to identify matching phrases and sentences. Some, like Turnitin, provide a percentage indicating how much of a student’s writing overlaps with existing works. However, due to potential false positives and the ambiguity surrounding intentional duplication, some educators have become hesitant to rely on these tools.

Now, the focus has shifted from merely detecting copied content to identifying AI-generated work. As students quickly adopt tools like ChatGPT, Google Gemini, and Microsoft Copilot, teachers are just as swiftly integrating AI detectors into their assessments. A survey revealed that 43 percent of teachers in grades six through twelve in the US regularly utilized AI detectors between 2024 and 2025. Some universities that already employed Turnitin found that the service automatically activated AI detection when it launched in 2023.

Unlike traditional plagiarism checkers, AI detectors such as GPTZero, Pangram, and Turnitin’s own system use AI models to assess whether a piece of writing is likely not human-generated. This method is arguably more complex than simply matching text online. According to GPTZero, these detectors analyze a text’s wording, rhythm, structure, and patterns in length and tone that may be indicative of AI authorship. This subjective evaluation can be less reliable, particularly for non-native English speakers. Despite this, Turnitin claims its AI detector incorrectly flags less than 1 percent of human-written content as AI, while Pangram asserts a false positive rate of just 1 in 10,000. GPTZero also reports a similarly low rate of misidentifying human content as AI.

How it’s going
Online, accusations of “sounding like AI” are becoming commonplace, and the availability of AI detection tools is intensifying this scrutiny. In notable instances, allegations of AI writing have had serious repercussions for individuals’ careers and reputations. Recently, a publisher dropped a $2 million book deal with author Jerry Falade over suspicions of AI use, which he strongly denies.

Thierry Rignol, a French national, sued Yale after a professor accused him of using AI for parts of his final exam, leading to a failing grade and a one-year suspension. The professor employed GPTZero to analyze Rignol’s writing, but the lawsuit contends that “AI surveillance and detection tools are known to unfairly target non-native English speakers” like him. In another case, a student at Adelphi University successfully sued the institution after a professor claimed he used AI for an essay. While the lawsuit doesn’t specify which AI tool was used, Adelphi has a licensing agreement with Turnitin.

A 2023 Stanford study indicated that AI detectors were more likely to misidentify essays by non-native English speakers as AI-generated compared to those by native speakers. Many services still argue their tools are effective with non-native text, but biases against neurodivergent writers have also been noted.

According to the University of California, Los Angeles, AI detection tools are designed to identify patterns that may suggest AI authorship, such as repetitive phrases, overly formal or informal language, and nonsensical wording. Some tools, like QuillBot, assess the “unpredictability” of text, as “AI tends to make the most ‘obvious’ or common language choices compared to human writing,” according to UCLA. They may also flag consistent sentence structures as another indicator of AI. However, these characteristics alone do not definitively indicate AI authorship, as some individuals may naturally have a writing style that exhibits these traits.

Despite Turnitin’s claims of low false positive rates, it acknowledges that its tool “may not always be accurate” and should not be the sole basis for taking action against a student. Grammarly cautions users to “never rely on the results of an AI detector alone,” while GPTZero states that “no AI detector can ever truly be 100% perfect.” OpenAI even discontinued its own AI writing detector in 2023 due to accuracy concerns.

Nonetheless, accusations of AI writing continue to circulate. Recently, Jack Osbourne, son of Ozzy Osbourne, publicly accused journalist Kat Tenbarge of using AI for an article, citing results from an AI detector as “proof.” Tenbarge has denied the claim, but Osbourne has not retracted his statement, leaving her to contend with online backlash.

Numerous instances of false accusations exist, affecting both writers and students, with many accusers overlooking the disclaimers associated with AI detection tools.

What happens next
The uncertainty surrounding AI detectors has prompted some educational institutions to abandon their use entirely. Yale, Johns Hopkins, Vanderbilt, Georgetown, and others have either disabled or limited AI detection tools. The Massachusetts Institute of Technology has also warned that “AI detectors don’t work.”

Rather than relying on these tools, many schools are encouraging educators to rethink their teaching methods. The University of Chicago suggests that instructors encourage students to slow down their reading, break up longer assignments, and reflect on their work. Stanford recommends in-class assessments, while MIT advises professors to allow students to disclose any AI assistance without facing penalties.

As AI becomes more integrated into both academic and everyday contexts, the efforts to distinguish between human and machine-generated content are intensifying. Some online platforms are only fueling suspicions about AI use. Substack has integrated Pangram into its app for scanning blogs for potential AI-generated content, while LinkedIn has introduced a “seems like AI slop” button for posts. This has ushered in a new era of skepticism, where readers constantly question the authenticity of what they consume, and human writers strive to avoid sounding like AI.

By the way
* The Authors Guild is assisting writers in preemptively addressing accusations by offering “Human Authored” certifications. There are also badges like Not by AI and Written by Human that users can attach to their work online.
* Wikipedia has created a guide to help editors identify AI writing, which includes looking for content that exaggerates a topic’s importance or provides superficial analysis. The site has also banned AI-generated articles.
Read this
* A fun quiz challenges you to compare pairs of passages and determine which one is AI-generated.
* Educators have shared their approaches to AI in schools, with one lecturer expressing concerns about the resources needed to make assignments AI-proof.
* The fanfiction community is grappling with its own challenges in identifying AI-generated stories.