
The flipped classroom is making a comeback in 2026. But not for the same reason it originally became popular. In the 2010s, flipping was about making class time more active. In 2026, it’s more urgent. Homework used to be where students did the hard thinking. Now AI can do it for them. Flipping brings that thinking back into class, where you can see it happen.
In the last week alone, three instructors told me they’d made the switch, though none of them called it a flipped classroom. They all described the same shift: basics out, application in. Exams, quizzes, and problem-solving are moving back into class, and teaching fundamental concepts is moving out. Flipping used to feel like a worthy but difficult redesign. Many of the instructors I speak with now see it as a necessity.
Why instructors are switching
One of them told me exactly where AI was hurting his students this semester. The week before his first test, his students upload the slides into AI and ask it to "teach me this in a week." On the first test, which covers the basics, that works: they get about a B. On the second, where they have to dig deeper and apply the principles, they slip to about a C. Once you take away the AI, he said, “they don't really know the material.” The exception is his A students, who still do the work. The ones who suffer are in the middle: students who used to pull their grade up are now slipping one instead.
MIT professor Eric Klopfer has a name for this: the "illusion of learning." Students turn in better homework but do worse on exams, because it feels like understanding until they have to apply it without AI. It's not exclusively an AI problem, but AI puts it on steroids. You feed lecture slides into an AI tool and tell it, “Help me learn this in under 30 minutes.” It spits out condensed information, asks you questions, and when you don’t understand something, it gives you easy answers. It feels like learning, but it skips the struggle that makes learning stick.
An instructor I spoke with saw it more starkly. He recently gave 15 MBA students an oral check on a valuation assignment they had already submitted, and none could explain their own work.
The research backs this up. In a randomized trial with 120 undergraduate business students, those who studied with unrestricted ChatGPT finished their assignment faster (3.2 hours vs. 5.8) but scored lower on a surprise test 45 days later (5.75 vs. 6.85 out of 10). In a separate study of nearly 1,000 high school math students, those using ChatGPT solved 48% more practice problems correctly but scored 17% worse on the exam. Worryingly, the students who used ChatGPT didn’t think they’d learned less.
So instructors are bringing that hard thinking back into class. Practice problems that used to be homework are now worked through on paper in class, where AI can’t do them in seconds. Tests and quizzes are moving back too, because class is where instructors can confidently measure what students have actually learned.
But that comes at a cost. If class time is now used for practice and assessment, when do students learn the fundamentals that used to be taught in class?
That's where the flip comes in. AI can do a good job of teaching the basics before class when it's set up to make students think, not just hand them answers. In a Harvard trial, students who learned introductory physics from a purpose-built AI tutor learned more than twice as much, in less time, as students in an active-learning class. The authors' own recommendation is to flip: use AI to teach introductory material before class, and spend class time on higher-order problem solving.
Assign readings, or give students a prompt for their AI tool of choice, and they can cover the basics before class. That frees up class time for the work AI can’t, or shouldn’t, do for them.
What instructors I’ve spoken with are doing
I analyzed calls the team and I have had with 99 instructors this year (I know, one more would have made this cleaner). Conversations that covered flipping the class went from about one in ten in 2025 to more than one in four this year. Of the 99, nearly half (48) now use class time for applied work like problems, cases, and Excel builds, and almost all of those have moved the basics outside class too. Another 24 are planning to make the switch. Most of them said AI was the reason. Here are some of the ways instructors are adapting:
Giving students a structured AI assignment. The instructor whose middle students were slipping wrote a four-page AI tutorial that students drill with before tests. On this semester's first test, he says every C grade came from students who hadn't done it.
Checking the prep at the door. Another instructor noticed that when students took quizzes online, "everybody was getting a 10." Now students do the videos and readings before class, then take a short paper quiz at the start of class, and he goes over whatever they struggled with. "It got people to really read or go through the videos before they came to class so we could have a better discussion."
Bringing the case work into the room. One instructor used to hand out valuation case studies as homework. Now, he says, "they'll just upload it, AI will do the work, and they'll turn it in." So he no longer sets homework. Students work through the cases in groups at their tables during class, and he walks the room.
Going "lecture-lite". A full flip isn't right for everyone. The most common pattern is lecture-lite: a short lecture, then the rest of the session on applied work. As one instructor put it: "I talk to them for maybe 40 minutes, and then they spend the next 30–45 minutes problem-solving, because that's better learning for them."
What I'd recommend
I'm a big fan of the flipped model: learn the fundamentals outside class, then work through the harder applications with an expert in the room.
Here's one way to start. Pick one topic coming up in the next few weeks.
Step 1: Move the basics out. Instead of teaching it in class, give students the slides and this prompt to use with their AI tool of choice:
"I'm a student preparing for class on [topic]. I've attached the slides. Don't summarize them for me. Teach me one concept at a time: ask me a question first, let me try to answer, and only then explain. If I get something wrong, give me a hint, not the answer. When we're done, give me three questions I should be able to answer without your help, and tell me which concepts I struggled with so I can bring them to class."
Want a more rigorous version? Here's a research-backed one to share with your students: Your AI Study Tutor.
Step 2: Check the prep at the door. Start class with a five-minute paper quiz on the topic, like the instructor above. It's harder to fake than an online quiz, and the results tell you what to go over. (Worried about grading all that paper? That's what BT’s free Paper Exam Grader is for.)
Step 3: Bring the application in. Spend the class time you've freed up on application problems. That's where the illusion of learning gets exposed, and where you're in the room to help.
Instructors didn't return to flipping because they suddenly had time to redesign their courses. They returned because the work that builds understanding stopped happening at home. Now it has to happen in the room.
If you've flipped, what are you doing with class time? If you haven't, what's holding you back?
Sources
Klopfer quote: "'It's the illusion of learning': How some educators say AI hurts students," The Washington Post, Sept 2026 (reprinted by the Anchorage Daily News): https://www.adn.com/nation-world/2026/09/22/its-the-illusion-of-learning-how-some-educators-say-ai-hurts-students/
Business student trial: Barcaui (2025), "ChatGPT as a cognitive crutch," Social Sciences & Humanities Open: https://www.sciencedirect.com/science/article/pii/S2590291125010186
High school math study: Bastani et al. (2025), "Generative AI without guardrails can harm learning," PNAS: https://doi.org/10.1073/pnas.2422633122
Harvard AI tutor trial: Kestin et al. (2025), "AI tutoring outperforms in-class active learning," Scientific Reports: https://pmc.ncbi.nlm.nih.gov/articles/PMC12179260/