Everyone in education, from K-12 teachers to university presidents, is well aware that AI is transforming the classroom. That presents all the challenges you’ve probably already heard of: students using ChatGPT to cheat, churning out papers and assignments without a second thought. But there’s also the more underreported development — teachers are deploying the technology to write lesson plans, make quizzes, and streamline administrative tasks, saving them hours of grunt work.
In the best-case scenario, AI promises to make teachers better at their jobs. And ultimately, if AI becomes the transformative force optimists hope it will, that will help students get smarter, becoming a tireless teaching aid and providing 24-hour tutoring assistance. That’s a big if, of course.
At the very least, the time saving element for teachers is real, and it’s a big deal. A recent survey from Gallup and the Walton Family Foundation found that 6 in 10 teachers used AI for their work in the 2023-2024 school year. Those that used AI weekly — about a third of the teachers surveyed — estimated it saved them about 6 hours each week, which, in the best of circumstances, could mean that’s 6 more hours of face time with students.
“This is not plugging students in front of computers, engaging with a chat bot,” Chris Agnew, director of the Generative AI in Education Hub at Stanford, said. “This is supporting teacher practice and then enabling this trained, experienced adult that’s in front of kids.”
Of course, giving teachers some time back doesn’t necessarily curb AI cheating. The savviest educators have clear guidelines for when AI can be used and when it can’t, as well as a good system in place for discussing the technology’s evolving role in school. After all, this is hardly the first time a new technology has swept into schools and upended old ways of doing things — educators used to worry about calculators in the classroom.
This also isn’t the first time a new technology has opened up a huge business opportunity for tech companies to reach young, inquiring minds and make a lot of money in the process. Google, for instance, now offers its Workspace for Education with Gemini built-in for up to $66 per teacher per month. In a school district of 500 teachers, that could easily add up to an extra $400,000 a year. For school districts that use a learning management system, like Canvas by Instructure, or an AI-powered tutor, like Khanmigo by Khan Academy, the cost of tech-centric education could keep growing.
“We went from the phase of, ‘Ban AI, it’s a cheating tool,’ to now, the majority of the market really is, ‘How do we leverage these tools in really productive ways?’” said Ryan Lufkin, vice president of global academic strategy at Instructure, whose Canvas software is used by half of North American college students and over a third of K-12 students.
What the classroom experience will look like in a decade, much further into the AI revolution, is anyone’s guess. In corporate America, companies are pouring billions of dollars into AI, hoping for transformative profits. So far, that’s not going great.
If you’re a parent, you might feel a bit powerless in this situation. Tech companies and school districts are making decisions that will impact your kid, who may or may not be using ChatGPT already to do their homework. But because we’re in the early days of this technology, now is the time to learn about how it works and what your school district is doing with it.
Beating the cheating problem
If you set aside the idea that large language models could reinvent the American education system — which is not great, by global standards — you might be curious about the ChatGPT cheating problem, especially if you’re a parent.
It’s hard to tell just how many students are cheating with robots. A Pew survey of teens found that 26 percent of middle and high school students were using ChatGPT — for both nefarious and less nefarious purposes — in 2024, a percentage that had doubled since 2023. Another study from 2024 that tracked high school students’ cheating from before and after ChatGPT’s release found no indications that it had “dramatically changed the prevalence of cheating.” Regardless, a New York magazine feature earlier this year declared that “ChatGPT has unraveled the entire academic project.”
Proposed solutions to the cheating problem, however serious it is, are kind of funny. As the use of ChatGPT has increased on college campuses, for instance, so have the sales of blue books, according to the Wall Street Journal. Students can’t use AI when they’re locked in a room with nothing but a pencil and paper, after all. Then there’s the call to bring back oral exams, including proposals to use video conferencing software to conduct hundreds of them at once. Researchers at the Georgia Institute of Technology even invented a platform for oral exams that, somewhat ironically, uses AI to grade the students. There are other creative workarounds, too, like requiring students to show their work by tracking changes in Google Docs or asking them to generate ChatGPT essays and then critique them.
Banning AI completely is increasingly unpopular. New York City Public Schools, the nation’s largest school district, banned ChatGPT not long after its release in 2022 and then lifted that ban a few months later. “The knee-jerk fear and risk overlooked the potential of generative AI to support students and teachers, as well as the reality that our students are participating in and will work in a world where understanding generative AI is crucial,” then-chancellor David Banks wrote in an op-ed. “While initial caution was justified, it has now evolved into an exploration and careful examination of this new technology’s power and risks.”
That exploration period seems to be ongoing for many K-12 schools. By the end of last year, the city’s comptroller, Brad Lander, called on the city’s Department of Education to pull nearly $2 million in funding for AI software, because it had not studied the efficacy of AI in the classroom.
The next edtech gold rush
Schools are nevertheless spending money on AI tools, whether they’re for teachers or for students. This represents just the latest raft of investment in education technology, or edtech. For the past four decades, putting computers into classrooms and screens in front of students has promised to transform learning. And for 40 years, it’s failed to fulfill that promise. Student performance has remained flat, while spending on edtech and training teachers how to use it has grown.
It’s unclear if AI can change this trend. Once you look beyond trying to stop students from using ChatGPT to cheat, you can get pretty creative with how AI might play a role in the classroom. You could imagine, for instance, that students will write fewer essays and might instead interact with a chatbot the way they’d talk to a human tutor. Khan Academy, a major edtech company, is piloting a chatbot it built with OpenAI called Khanmigo in 266 school districts nationwide. Khan Academy founder Sal Khan recently told Anderson Cooper that his dream is to give every student a private tutor. Khanmigo currently costs $4 per month per student.
What’s more promising in the near future is giving teachers access to AI that can lead to new classroom experiences. Aside from its private tutor powers, Khanmigo can help create lesson plans and then integrate the chatbot into them, according to Kristen DiCerbo, chief learning officer at Khan Academy. She explained a scenario to me in which several Khanmigo agents essentially worked like teacher’s assistants, checking in on groups of students during a lesson. “We think of it as like a force multiplier for the teacher, giving them just a little more reach in terms of what they can get done in the classroom,” DiCerbo said.
Aside from powering tools like this, OpenAI recently announced an education effort of its own in ChatGPT called study mode. This effectively turns ChatGPT into a tutor that replies to questions with more questions rather than answers. This is in addition to ChatGPT Edu, which launched last year and offers a version of ChatGPT built just for universities at a discount. Google is similarly marketing its Gemini Pro plan to students, who can currently get one year for free. Anthropic is selling a version of its Claude chatbot to universities, too. All of these education-specific products work a lot like the consumer versions but don’t train their models on student data.
That all sounds good in theory, and it all costs money. It goes without saying that schools with more resources will be able to take better advantage of these new AI tools, possibly improving teachers’ lives and student performance.
“Technology is not and never has been a silver bullet to address some of these more structural issues that exist in our education system,” said Robbie Torney, senior director of AI programs at Common Sense Media.
This is assuming that AI in education actually delivers the desired results, which would defy the decades-long trend in edtech. Despite initiatives that date back to the ’90s to give schools cheap and easy access to the internet, a quarter of the school districts in the US don’t even have broadband that’s fast enough to support some of these applications. It’s hard to have an education revolution when the page won’t load.
So, for a number of reasons, chatbots won’t be replacing teachers any time soon. More teachers may enlist AI to mix up their lesson plans, and students will inevitably try to find high-tech ways to get homework help. A chatbot that refuses to tell them answers might be their best hope.
A version of this story was also published in the User Friendly newsletter. Sign up here so you don’t miss the next one!
The post Cheating is the least AI can do appeared first on Vox.
