Blue books made a comeback in 2025. In an effort to prevent students from feeding final essay prompts into ChatGPT, some professors asked their students to sit down and write in-person in the lined, sky-blue booklets that served as the college standard for written assessments in the pre-laptop era. But it may not be the foolproof way to prevent AI-assisted cheating that faculty are looking for: Meta now offers Ray-Ban glasses with a built-in AI assistant that sees what the wearer sees and can communicate silently and privately via an in-lens display.
“What is to stop someone from sitting in the back of a classroom and whispering into their glasses to say, ‘Hey, I need help with solving this problem,’” said Luke Hobson, an assistant director of instructional design at MIT and lecturer at University of Miami’s School of Education and Human Development. “Every time I see someone saying, ‘Blue books are the future,’ I’m like, ‘So are we going to ban students from wearing glasses?’”
And glasses are just one of the newest AI gadgets. Students can also talk to their smartwatches, rings or other AI-infused wearables. Though a long way from widespread use, Neuralink began trials this year for its brain-computer interface, which is designed to decode neural activity so that users can control a computer or smartphone simply by thinking about it.
“Predominantly I teach online courses, and … these AI companies have now created these different types of browsers that have agents embedded inside of them,” Hobson said. “They can literally take online courses just by prompting them to do so. That’s not the same as copy-and-pasting inside ChatGPT.”
Initial conversations around AI and assessments focused on cheating detection and enforcement, and—led by GPTZero—technology companies flooded the market with programs promising to detect whether a piece of writing is AI-generated. But experts agree that it’s impossible to be completely certain whether a student used AI—the technology is just too good. Instead, instructors should focus on re-engaging students, using class time well and introducing a variety of assessment types.
“For so long, online courses have been the same old, same old—essays and multiple choice questions. Now it’s like, ‘Okay, let’s elevate this. Let’s really make this into a whole new type of learning experience to make it better,’” said Hobson, who has taught courses on instructional design, teaching and learning, and business. He recently laid out five AI-resistant assessment options in a LinkedIn post, some of which he’s implemented in his own classroom.
One assignment he’s used is recorded journals, which require students to regularly record and upload five-minute videos of themselves talking about what they’ve learned in class, how it connects to their past experiences and how they might use it in the future. He has also asked his instructional design students to record interviews with professionals, listen to each other’s recordings and reflect on those conversations.
Oral assessments—either stand-alone or accompanying an essay—are also becoming more popular. The ancient assessment asks students to sit before their professor and talk about what they’ve learned, answer questions in-person or explain their writing process. Students are evaluated in real-time, making it difficult, if not currently impossible, to use artificial intelligence to cheat.
Another strategy Hobson recommends is asking students to critique AI. Seeing the technology’s limited existing capabilities has helped assuage some students’ fears about being immediately replaced in the workforce, he said. Beth Rochefort, an AI adoption consultant and director of digital transformation at Northeastern University, has largely embraced using AI in her education and project management classes, but asks her students to explain how and why they use it.
“Usually in presentations I’ll ask, ‘Okay, so tell me how you used AI in this project,’” Rochefort said. “What did you do? What was helpful, and what did you have to go back and check and why?”
ChatGPT Can’t Put Up Drywall
One type of assessment remains largely AI-resistant: hands-on projects, said Leon Furze, a Ph.D. candidate studying generative AI at Deakin University in Australia and author of the book Practical AI Strategies. Artificial intelligence has yet to be very helpful in the physical world.
“There are many, many disciplines that don’t rely on essay-based assessments. Anything which is practical or vocational education—those kinds of fields are technically AI-proof,” Furze said. “I can’t imagine ChatGPT helping anybody to put up a drywall or do their electrical engineering.”
Carlo Rotella, a professor of American studies, English and journalism at Boston College and author of the book What Can I Get Out of This? purposefully does not ban students from using AI. It’s not a policy he could fairly enforce, he said, because he recognizes that some students may be submitting AI-completed assignments undetected. “I explain to my students why it’s a waste of their time and mine. I explain that they’re paying $5 a minute for classes at Boston College, and to spend that time practicing to be replaceable by AI is a complete waste of their money and time, and my time.”
Rotella is especially focused on what happens in the classroom.
“The 2,000 minutes we spend together in class is the main event,” Rotella said. “If you’re going to make that the main event, then you build the structure of the class around that. That means, in my case … I expect everyone to speak up at every class meeting.”
He does not allow devices in class; students must bring hard copies of the reading. Much of the content students are tested on comes from in-class discussions, something an AI assistant who wasn’t present wouldn’t know anything about. He’ll occasionally give a quiz—an idea he learned from University of California, Berkeley, English professor Scott Saul—that asks students to recall small, memorable details that an AI-generated summary of a book would not include.
Furze advocates this type of student-and-classroom focused approach. Getting to know one’s students is a great way to understand how they’re using AI, he said.
“If we can reduce cohort sizes, spend more time conversing with students, and more time in dialogue with students building those relationships, that makes it easier to spot when a student suddenly changes voice and style entirely and sounds like ChatGPT,” Furze said. “But also, I think it makes it less likely that the students might be inclined to cheat their way through in the first place.”
Experts also agree that instructors must remind their students that learning requires practice. As Rotella tells his students, the point of almost all coursework is the act of doing it.
“The entire point of this class is the labor, so a labor-saving device would be beside the point,” Rotella said. “It’s like joining the track team and doing your laps on an electric scooter. You went around the track. Congratulations.”