Students Are Increasingly Using AI to Cheat — And Getting Away With It – entrepreneur.com

You’re viewing the US website. Looking for the Site edition?
It’s harder than ever to detect AI-written text and easier than ever to churn it out.
Copied to clipboard
One look at social media makes it clear — it is easier than ever to cheat with AI and get away with it, leaving students with an ethical challenge.
Students are increasingly using AI writing tools to complete essays while slipping past school detection systems, The New York Times recently reported. Short videos on TikTok and YouTube walk students through step-by-step instructions for generating essays with chatbots, then adjusting that text so it appears as though a human being wrote it. 
These videos advise students to treat AI as a ghostwriter rather than just copying and pasting from a chatbot. In other words, students run prompts through ChatGPT-style tools, then process the output through additional apps to disguise its origin before submitting it as their own. 
A key tactic involves “humanizer” tools that rewrite AI text to look more like something a student would write. These services add variation in sentence length, change up vocabulary and sometimes add mild errors or slang to break the smooth flow of words that many detectors look for. 
“Autotyper” tools go further by simulating the physical act of writing. They slowly type the essay into Google Docs or other platforms in real time, complete with pauses, edits and occasional backspacing, leaving behind a version history that looks like a genuine draft rather than a pasted block of text. Some even sprinkle in intentionally misspelled words.
According to the Times, AI detectors are unreliable and often produce both false positives and false negatives. They can label polished prose written by a human being as AI-generated, while simultaneously allowing heavily edited AI text or “humanized” content to slip through the cracks. 
Detectors typically scrutinize text for statistical patterns, such as predictable word choices and familiar sentence structures. However, once students paraphrase, add in personal details or run their essays through humanizers, detection accuracy drops. Because of this, instructors are hesitant to treat detector scores as hard proof of cheating, per the Times. 
At Harvard, the conversation around AI cheating is especially charged. The university has imposed strict limits on AI for tasks like admissions essays, banning generative tools. 
Jenny Ng, a Harvard student, told the Times that despite the university’s best efforts, AI use remains widespread, even as students experience “a level of shame” about relying on AI tools. She said professors have reacted to AI by administering pen-and-paper tests and weighing these exams more heavily than other assignments. 
Ng recently created a TikTok video for Grammarly, demonstrating how she used the AI tool to generate a study guide for a statistics test. According to the Times, professors are increasingly worried about Grammarly, which has been live for 17 years as a writing assistant that analyzes sentence structure, vocabulary and tone. Grammarly now allows students to generate writing and humanize text with AI. It also detects phrases that could trigger AI detectors. 
Grammarly “is a suite of tools that will do everything for you,” George Cusack, director of AI academic initiatives at Carleton College, told the Times. “It’s kind of shocking.”
One look at social media makes it clear — it is easier than ever to cheat with AI and get away with it, leaving students with an ethical challenge.
Students are increasingly using AI writing tools to complete essays while slipping past school detection systems, The New York Times recently reported. Short videos on TikTok and YouTube walk students through step-by-step instructions for generating essays with chatbots, then adjusting that text so it appears as though a human being wrote it. 
These videos advise students to treat AI as a ghostwriter rather than just copying and pasting from a chatbot. In other words, students run prompts through ChatGPT-style tools, then process the output through additional apps to disguise its origin before submitting it as their own. 

source