Our academic system has combusted. For decades we relied on written output and performance to both measure and push student capability. Generation was the vehicle of learning. Now that model simply doesn't work. The majority of students now admit to using AI to complete assignments. Many will say they are using it to study and understand complex ideas, but the data contradicts this. Scores on take-home work have improved significantly, yet performance on tests has declined. If the goal is for students to become more capable, any teacher in the country can tell you that the model is failing. Where do we go from here?
Rank of "preventing cheating" among faculty instructional concerns, 2023 (up from #10 in 2022)
Faculty concern about cheating jumped to the top spot within a single year of ChatGPT's release.
Tyton Partners — Tyton Partners — Time for Class 2023, 2023
Self-reported high-school cheating rate, essentially flat before and after ChatGPT's release
Stanford / Challenge Success — Lee, Pope, Miles & Zárate — Cheating in the age of generative AI, Computers and Education: AI (2024), 2024
Faculty perception, three years later
Faculty who say cheating has increased
Faculty who say it increased "a lot"
Faculty who personally handled an integrity case
Elon University / AAC&U — Elon University / AAC&U — The AI Challenge faculty survey (2026), January 2026
AI-writing flags in take-home work
Turnitin — Turnitin — AI writing detection data releases (2023–2026), 2023–2026
What K-12 classrooms report
Center for Democracy & Technology — Center for Democracy & Technology — K-12 teacher and student AI surveys (2024, 2025), 2024–2025
If we want to answer this question, we have to start back at the foundation. What is it that we are trying to accomplish in the first place? What is education? Intuitively, we know education is meant to build student capacity, but capacity towards what? At the moment, our target is performance on standardized testing. This is not because educators believe this is the right metric, but because standardized scores translate to teacher salaries, maintenance, student aid, and everything else that runs the institution. So, we hold on to this paradigm.
As has become apparent, this pedagogical method is not compatible with AI. We train students for tests by giving them mini take-home versions of tests called homework. Outside of the classroom, ChatGPT, Claude, or Gemini can complete these assignments at 100% accuracy while providing full explanations. There is no real information that you can ask a student to put on a homework assignment to prove they did not use AI assistance. AI can generate whatever the student asks it to.
Teens using ChatGPT for schoolwork
2023
2024
Pew Research Center — Pew Research Center — About a Quarter of U.S. Teens Have Used ChatGPT for Schoolwork (2025), January 2025
Middle & high schoolers using homework AI
May 2025
Dec 2025
RAND Corporation — RAND — American Youth Panel, AI for Homework (Dec. 2025 wave), December 2025
U.S. high schoolers using any GenAI tool
Jan 2025
May 2025
College Board — College Board — U.S. High School Students' Use of Generative Artificial Intelligence (2025), 2025
UK undergrads using AI in assessments
2024
2025
Higher Education Policy Institute — HEPI/Kortext — Student Generative AI Survey 2025, 2025
So what else do we aim for? We knew the ideal answer long before AI: allow the learning to be driven by curiosity and build on each student's strengths; customize learning paths based on individual needs; build a joy for learning. But these goals are impractical. We don't have enough resources to focus on every student. So, instead we must structure our schools as an assembly-line factory.
This is wrong. We know a lot about how the mind learns — what works and what doesn't. The problem of the past was scalability, but much of this can be automated by AI systems. Companies like Alpha School and Math Academy demonstrate that effective learning is a science, and when put in the right environment, almost every child is an exceptional learner. These tools will only become better and cheaper. Presumably, every student will have access to personalized tutors. Academic growth will be greatly accelerated while student understanding will be more accurately measured. If our goal is simply to build student capacity, this sounds like the optimal design. It is faster, cheaper, and, if done right, more equitable. The "AI problem" we now face is not driven by irresponsible students. It is a product of our own unwillingness to adapt. The answer is to fight fire with fire! But why does this feel wrong?
In our heart we know that capacity is not the end-goal of education. We are not in the business of training machines. The skills we teach like math, science, english, and history are cognitive tools for observing and changing the world. But what good is giving a kid a hammer if they have no vision for what to build? We can train students to be experts at every academic skill, but these are truly meaningless without application. We need to teach students what it means to participate in community, in society. How to build on their passions and make meaningful contributions to themselves and the people they love. This is a human skill. It involves interacting with and empathizing with other people. Skills that can only be learned through experience.
As educators, our roles will change. Instead of raising test-takers, we will be raising creators. We must ask ourselves, and the students, what kind of world do we want to live in? As we chase this ever-evolving vision in our minds, we can, at the same time, build it here and now. The world ahead looks bright. Humans can never be replaced, but the fear of replacement will have consequences. Who do you want leading the future? Do you believe in your students? Do you believe in yourself? If not, who are you putting your trust in? We have entered a period of rapid change, and we each decide the direction.
Sources
- Tyton Partners — Tyton Partners — Time for Class 2023, 2023
- Stanford / Challenge Success — Lee, Pope, Miles & Zárate — Cheating in the age of generative AI, Computers and Education: AI (2024), 2024
- Elon University / AAC&U — Elon University / AAC&U — The AI Challenge faculty survey (2026), January 2026
- Turnitin — Turnitin — AI writing detection data releases (2023–2026), 2023–2026
- Center for Democracy & Technology — Center for Democracy & Technology — K-12 teacher and student AI surveys (2024, 2025), 2024–2025
- Pew Research Center — Pew Research Center — About a Quarter of U.S. Teens Have Used ChatGPT for Schoolwork (2025), January 2025
- RAND Corporation — RAND — American Youth Panel, AI for Homework (Dec. 2025 wave), December 2025
- College Board — College Board — U.S. High School Students' Use of Generative Artificial Intelligence (2025), 2025
- Higher Education Policy Institute — HEPI/Kortext — Student Generative AI Survey 2025, 2025