Schools Are Banning AI. Students Are Using It Anyway.

Teens are using AI for schoolwork regardless of school policies and parents are unaware teens are using AI as much as they are. Teens and adults are both wary of AI, with power users more positive.

New research from the University of Southern California, Center for Applied Research in Education with the support of the Marshall School Neely Center.

Schools across the nation are grappling with what to do about Artificial Intelligence. As the technology rapidly unfolds, questions abound – what, if anything, should students be learning, using, and doing with AI? And how should teachers be learning, using, and integrating AI technologies into their instruction?

In November 2025, we surveyed a nationally representative sample of over 1,000 adults through the Understanding America Study, and the 348 teens in their households, to gain a national portrait of AI school policies, how students are using AI in school, and parents’ and teens’ attitudes and opinions about AI in education. (Though we sometimes use “parents” when describing results, we note that respondents are parents about two-thirds of the time; the others are grandparents, adult siblings, and other adult family members). Lessons learned can inform the conversation around AI-related practices and policies in schools going forward, though it seems we are far from an emerging consensus on key issues.

School policies are not yet clearly established

Roughly one-third of responding teens told us that their school has an AI-ban, with students not allowed to use AI at all. AI bans were reported as slightly more common among middle schoolers compared to high schoolers. Another half of the sample reported their school has rules specifying when students can/can’t use AI rather than a complete ban. Over 50% also said that different teachers have different rules, which was slightly more common in high school. Parents mostly don’t know what the rules are in their children’s schools, with 55% reporting they are “unsure” what the rules are (52% of parents of teens specifically).

Teens are using AI for schoolwork regardless of school policies

Despite a possible reluctance to admit to doing so, two-thirds of teens surveyed reported they use AI for schoolwork, similar to Pew’s recently released survey results. And this number is increasing quickly, as only 55% said they were using AI for schoolwork as of last summer. Roughly one-third said they never use AI for schoolwork, another 40% use it once or twice per week, and the remaining 26% use it multiple times per week or daily. These usage patterns do not vary meaningfully based on a school’s policy, with just as much use in schools with bans as in those without.

We removed a layer of social-desirability pressure (or fear of being ‘caught’) by asking teens how often their friends use AI for schoolwork, adding “even when they are not supposed to.” Over 40% said most or all do so; this total percent increased to 70% when we included those who said “some” of their friends are doing so. AI use for schoolwork is clearly already widespread among teens, including in schools that have policies against it.

Parents are unaware teens are using AI as much as they are

Despite these high reported use levels, only 9% of teens’ parents think their teens are regularly using AI for schoolwork. In fact, over the past two years, roughly half of all parents we’ve surveyed have consistently reported that their child has never used AI at all, despite rapid advances in the technology, sustained media coverage, widespread commercialization, and its growing presence in everyday conversation.

Moving forward, it is essential that schools and districts develop a clear policy and communicate it to students, teachers, and families, acknowledging the evolving nature of the technology and the potential for the guidelines or rules to change. State leaders could help by crafting model policies that districts can adopt or modify.

Average teen use may (at least sometimes) be reasonably productive

Acknowledging that social desirability bias possibly influenced teens’ responses, the primary reasons teens say they use AI were potentially productive. Over half said one of the top three reasons they use AI is to search for information, research a topic, or answer a question. This was by far the most often selected use, though it is unknown whether teens understand the potential pitfalls of trusting the results AI provides. Approximately one-third selected brainstorming ideas for a project, helping to problem-solve, and studying for a test – uses that even teachers might support. Just one in five selected, “Generating writing for homework” (the item most closely aligned to “cheating” without using that term) among their top three uses.

The prevalence of these different uses mostly did not vary based on school policy. Though there were fewer teens in schools with bans using AI to search for information, edit their own writing, or just see what it can do, there were relatively more teens in schools with bans using AI to help problem-solve, talk about a personal issue, or do something creative. Using AI to study for a test, and generate writing for homework did not vary based on school policy.

Both teens and adults are wary of AI’s impacts, or don’t expect any impact at all

On average, teens and adults both expressed more negative or ambivalent attitudes towards AI’s potential impacts than positive ones, with teens’ perceptions slightly more negative than adults’. A minority of teens believe AI will have a positive impact on their lives across a range of areas (green portion of each bar below, top panel), including job opportunities, education opportunities, and individuals’ ability to complete creative or intellectual tasks. Even fewer believe AI will positively impact individuals’ ability to develop meaningful relationships. In most areas, teens were split between negative (blue) and no impact (orange) response options. Adults displayed a slightly more optimistic outlook, particularly for job opportunities, for which almost twice as many adults than teens believe AI’s impact will be positive (green bars in lower panel).

Most teens believe AI will have no impact (orange bars) on a host of skills we asked about, including their curiosity/desire to learn, imagination/creativity, social skills, and motivation. But twice as many believed the impact would be negative (blue) as opposed to positive (green) for social skills, and more than three times as many believed impacts would be negative versus positive for attention span and their ability to think for themselves.

On the other hand, more teens thought AI’s impacts would be positive than negative for the ability to complete school tasks, study habits, and curiosity/desire to learn, though still more believe it will have no impact.

As an overall assessment, teens were modestly more wary about AI than parents – with 55% agreeing or strongly agreeing that AI causes more harm than good, compared to 48% of adults. Perhaps unsurprisingly, given these overall concerns, when asked to choose between an AI or human option for a range of services including a driver, lawyer, doctor, teacher, etc., 74% to 95% of both teens and adults chose the human option – tutors and financial advisors were the most likely to elicit an AI selection, but topped out at 26% of teens choosing AI (21% among adults).

Most teens do not believe AI changes the value of college or career fields

Though it seems reasonable that teens’ ideas about college and career prospects might be impacted by AI, as of spring 2026 a majority report that AI does not change the value of four-year college (53%), their likelihood of pursuing STEM careers (56%), or the field they think they will want to pursue (69%). Similarly, 69% of parents believe AI is not impacting teen interest in attending college, though a plurality (44%) believe AI is negatively impacting the quality of education.

Overall results mask variation in beliefs: AI “power users” have higher hopes; AI “never-users” have higher concerns

Anticipated AI impacts were quite different when we considered individuals’ AI usage. We categorized respondents into three different usage levels: 1) power users - if they use AI more than two times per week), 2) light users - if they reported using AI only once or twice or less than once per week, and 3) never users. Though most teens are skeptical of AI’s benefits, the percentage of teens optimistic about AI’s impacts was two to three times higher among power users than among never users across every skill and area we asked about. In the areas related to school and knowledge, five to six times more power users than never users expect AI impacts to be positive (e.g., 60% compared to 10% for ‘school tasks’ and 54% to 10% for study habits). Perhaps reasonably, but not obviously, users have a better understanding for the ways in which AI can be harnessed than non-users.

While there was no area in which fewer power users had an optimistic outlook compared to never users, a majority of all users are not optimistic in the areas of developing personal relationships, thinking for oneself, attention span, and to a lesser extent, social skills. These may be areas where even AI users seem attuned to AI’s potential harms.

Similarly, while almost two-thirds (65%) of never users believe AI will cause more harm than good, this drops to less than half (42%) of power users. Two-to-three times as many power users than never users would choose an AI option for a tutor, a financial advisor, a customer service representative, a driver, a teacher, and even for a lawyer, with almost half of power users choosing an AI over human tutoring.

These basic statistics suggest more experience and familiarity with AI seems associated with more positive views on AI, though it is also possible that those with more favorable attitudes toward the technology are more likely to try it and end up using it regularly.

Notably, teens’ usage levels were not as closely related to their feelings about their post-secondary education – most did not feel that AI changes how important they think school is (47%-49% across user categories), their interest in a STEM field (55%-63%), or the career field they want to pursue (64%-75%).

It is too early to know AI’s consequences for students and for education in general

We know that students are using AI and doing so more than adults think they are. We don’t really know if on balance, what they are doing is productive and positive, or reductive and harmful. There have been horrific incidents where teens have taken their own lives, or caused harm to others, based on an AI chatbots’ “advice.” And at least some students are inappropriately using AI to complete schoolwork in ways that bypass meaningful skill building, though to what extent is not really known. Critical thinking skills seem to be at greatest risk. Already a challenge in today’s public-school systems, even AI power users are wary of AI’s potential negative influence on developing needed critical thinking skills.

On the other hand, at least some teen AI use may be productive (or at least harmless), as they experiment with what the technology can do. As some of our power users perceive, AI can enhance a students’ creativity, imagination, improve study habits and help with schoolwork. For some students, being able to use AI in productive ways in school could improve their motivation.

One thing is certain – today’s students will be tomorrow’s engineers, CEOs, policy makers, business owners, software developers, politicians, etc. It is critical for students to become educated about AI’s potential risks as they are using it with or without adults’ knowledge and/or awareness. At a minimum, these results illustrate that we can’t simply ban AI and expect that students won’t use it. Instead, we need policies that promote safe and responsible use, with guidelines that have the flexibility to evolve as the evidence accumulates on AI’s impacts on the developing brain, on skill building, and on society as a whole.

*Analytic note: We also analyzed results separately limiting the sample to only parent-teen dyads from the same households (224 households with an adult respondent and 249 teen respondents). Results were not substantively different than those reported here.