The 63% Problem: Teens Who Get Answers From AI

Common Sense Media surveyed more than 1,000 U.S. teens in August 2026 and found that among those using AI for schoolwork, 63% get an answer to their schoolwork from AI. The breakdown is the interesting part: 25% use it as-is, 31% rewrite it to sound like them, 35% improve it with what they know. Notice that the last two groups are working. They’re just working on the wrong end of the problem. This teens AI answers habit isn’t a character flaw, it’s a habit loop with a very short reward delay, and habit science has a specific answer for that.

Key Takeaways

• Common Sense Media (Aug 18, 2026): 70% of teens use AI for schoolwork; of those, 63% get an answer, split 25% as-is, 31% rewritten to sound like them, 35% improved with what they know.
• The same survey: 38% say AI leads them to generate fewer of their own ideas, and 39% say they feel like they’re missing out on learning when they use it. The kids know.
• Wood and Rünger (2016, Annual Review of Psychology) describe habits as “the efficient, default mode of response,” formed by repeating the same response in a given context. Cue, not willpower.
• Lally et al. (2010) found a median of 66 days to maximum automaticity, ranging from 18 to 254 days. Plan for months.
• The workable intervention is changing the cue: what happens in the four seconds between “I’m stuck” and typing.

What the 63% number actually describes

Common Sense Media’s “Teens in the AI Era: Schoolwork and Skills That Matter,” released August 18, 2026, surveyed a nationally representative sample of more than 1,000 U.S. teens. Seventy percent report using AI for schoolwork. Of those users, 63% “get an answer to their schoolwork from AI,” and the report splits that into three behaviors: “use it as-is (25%), rewrite it to sound like them (31%), improve it with what they know (35%).”

Two other numbers in the same report deserve attention because they complicate the easy story. Thirty-eight percent say having AI “leads them to generate fewer of their own ideas.” Thirty-nine percent say they “feel like they’re missing out on learning” when they use it.

That’s not the profile of kids who think they’re getting away with something. It’s the profile of kids caught in a loop they can describe and can’t exit. Which is precisely what a habit is.

And the outcome data lines up. PISA 2025, released September 8, 2026, found students who never or almost never used AI to draft writing assignments scored 509 in science, versus 481 for near-daily drafters. Bastani, Bastani, Sungu, Ge, Kabakcı, and Mariman (2025, PNAS) found that in a field experiment with about a thousand high school math students, those using an unguarded ChatGPT-style tutor performed 17% worse than students with no AI access once the tool was removed.

Why this is a habit, not a decision

Wood and Rünger (2016) in the Annual Review of Psychology give the cleanest description: “habits form as people pursue goals by repeating the same responses in a given context,” and habits are “the efficient, default mode of response.” They also note that people “tend to infer from the frequency of habit performance that the behavior must have been intended,” which explains the defensive reaction when you ask a teen why they did it. They’ll construct a reason after the fact.

The structure of the AI-answer habit is unusually strong for three reasons.

The cue is sharp. “I don’t know how to do this” is an unmistakable internal signal, and it arrives at a predictable time of day.

The reward is immediate and total. Not partial relief. The problem disappears in four seconds. Almost no behavior in a teenager’s life has that reward profile.

The cost is delayed and invisible. The 17% deficit shows up weeks later on a test, by which point no one connects it to a Tuesday night.

Against that, “try harder” is not a plan. What changes habits, per the research, is altering the context and the cue, not increasing motivation. Lally, van Jaarsveld, Potts, and Wardle (2010, European Journal of Social Psychology) had 96 volunteers perform a chosen daily behavior in the same context for 12 weeks and found a median 66 days to maximum automaticity, with individual times ranging from 18 to 254 days. Repetition in a consistent context was largely sufficient. That’s the lever.

The habit-stage swap table

Each row targets one stage of the loop. Do not try all six at once; pick one and run it for three weeks.

Habit stage	What’s happening now	The swap	Why it works
Cue: “I’m stuck”	Hand reaches for phone within seconds	Write the stuck-point on paper in one sentence first	Inserts a 20-second delay and converts vague confusion into a specific question
Craving: wanting it gone	Anticipation of instant relief	Set a 10-minute timer before any AI use is allowed	Reward delay is the single most effective structural change
Response: typing the task	”Solve this / write this”	Same problem, study-mode phrasing: attempt first, ask for a question back	Same tool, different mode. Bastani’s guarded tutor avoided the 17% penalty
Reward: problem gone	Assignment complete, nothing retained	End with one self-test question, out loud, device down	Roediger and Karpicke (2006): testing beats restudying for retention
Context	Phone on desk, chat tab always open	Homework happens in a different room from the phone; one tab	Wood and Rünger: change the context, not the willpower
Identity	”I’m bad at this, AI is better”	Keep a running list of problems they solved unaided	Rebuilds the self-belief the habit eroded

The row that produces the most change per unit of effort is the cue row. Writing the stuck-point in one sentence, on paper, before touching the device, does two things: it inserts a delay, and it forces the confusion into words. Very often the act of writing “I don’t know whether to divide or multiply here” answers the question, and if it doesn’t, it produces a much better prompt.

Running the swap without a fight

Show them their own numbers first

Don’t open with a rule. Open with the Common Sense findings, especially the two that describe their experience: 38% generating fewer of their own ideas, 39% feeling like they’re missing out on learning. Ask if that matches. Most teens say yes, and the conversation stops being about compliance.

Pick one row and run it for three weeks

Not six changes. One. The Lally data says habit change takes a median of 66 days with enormous variance, so a three-week trial isn’t about installing the new habit; it’s about collecting evidence that the swap is tolerable. If they can live with it for three weeks, it has a chance.

Make the ten-minute timer the default, not a punishment

The timer says: you can absolutely use AI on this, in ten minutes. That framing matters. It’s not a ban, it’s a sequence, and it preserves their autonomy while breaking the four-second reward loop. Many kids solve the problem during the ten minutes, which is the best possible outcome because they experience the win.

Move the phone, not the kid

Context beats intent. If homework happens at the kitchen table and the phone lives on the counter, the reach for it becomes a decision instead of a reflex. That’s the entire mechanism, and it costs nothing.

Use the oral check as the reward step

The University of Chicago Law School’s July 9, 2026 AI strategy required students to defend research papers orally, with questions probing “the paper’s reasoning and argument implications,” instead of using detection software. The home version, two minutes of “walk me through this,” closes the loop by making understanding rather than completion the thing that ends homework.

What not to do

Don’t make it about honesty. The 31% who rewrite it to sound like them and the 35% who improve it already know it’s a gray area; framing it as cheating starts an argument about definitions instead of a conversation about learning. And don’t ban AI outright at home: Common Sense found that when schools blocked tools, 44% of teens hit a block and 59% of those simply switched to a personal device.

What to Watch For Over the Next 3 Months

• Week 4: Is the paper step still happening? This is the first thing to quietly disappear, and it’s the highest-value item.
• Month 2 red flags: They’ve started writing the stuck-point and then pasting the whole problem anyway. Also watch for the timer being set and ignored.
• Month 3 self-check: Look at the unaided-problems list. If it’s empty, the swap didn’t take. If it has ten entries, the identity row is working.

Frequently Asked Questions

What does the 63% figure actually measure?

Common Sense Media’s August 2026 survey of more than 1,000 U.S. teens found that among teens using AI for schoolwork, 63% get an answer to their schoolwork from AI: 25% use it as-is, 31% rewrite it to sound like them, and 35% improve it with what they know.

Is rewriting AI output better than copying it?

Slightly, in that the student processes the material. But the generative step, deciding what to say, is where most of the learning happens, and rewriting skips it. PISA 2025 found students who rarely used AI to draft writing scored 509 in science versus 481 for near-daily drafters.

How long does it take to change this habit?

Lally et al. (2010) found a median of 66 days to reach maximum automaticity for a new behavior, with individual times from 18 to 254 days. Plan in months, not weeks, and expect the first three weeks to feel like nothing is happening.

Why doesn’t just telling them to stop work?

Because habits are, per Wood and Rünger (2016), “the efficient, default mode of response” triggered by context and cues rather than deliberate choice. Interventions that change the cue and the context outperform interventions that try to increase motivation.

What’s the single best change?

Writing the stuck-point in one sentence, on paper, before touching a device. It adds a delay, converts vague confusion into a specific question, and often solves the problem outright. Second best: putting the phone in another room during homework.

Should I ban AI at home instead?

The evidence argues against it. PISA 2025 found moderate AI users outscored both non-users and near-daily users. And Common Sense found that when schools blocked AI, 44% of teens encountered a block and 59% of those switched to a personal device. Restriction relocates the behavior.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. Common Sense Media. (Aug 18, 2026). “Teens in the AI Era: Schoolwork and Skills That Matter.” https://www.commonsensemedia.org/research/teens-in-the-ai-era-schoolwork-and-skills-that-matter
2. Wood, W., & Rünger, D. (2016). “Psychology of Habit.” Annual Review of Psychology, 67, 289–314. https://doi.org/10.1146/annurev-psych-122414-033417
3. Lally, P., van Jaarsveld, C. H. M., Potts, H. W. W., & Wardle, J. (2010). “How are habits formed: Modelling habit formation in the real world.” European Journal of Social Psychology, 40(6). https://onlinelibrary.wiley.com/doi/10.1002/ejsp.674
4. Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). “Generative AI without guardrails can harm learning: Evidence from high school mathematics.” PNAS, 122(26). https://doi.org/10.1073/pnas.2422633122
5. Roediger, H. L., III, & Karpicke, J. D. (2006). “Test-enhanced learning: Taking memory tests improves long-term retention.” Psychological Science, 17(3). https://doi.org/10.1111/j.1467-9280.2006.01693.x
6. OECD. (Sept 8, 2026). “PISA 2025 Results (Volume I).” https://www.oecd.org/en/publications/pisa-2025-results-volume-i_73451bc5-en.html
7. Fortune. (Sept 11, 2026). “Students are using AI in school and test scores are down.” https://fortune.com/2026/09/11/students-ai-school-test-scores-moderate-use-vs-daily-users/
8. University of Chicago Law School. (July 9, 2026). “Rethinking Legal Education in the AI Era.” https://www.law.uchicago.edu/news/ai-strategy-statement
9. NPR/Ipsos. (June 5, 2026). “Poll: Teachers worry AI is impacting students’ critical thinking.” https://www.npr.org/2026/06/05/nx-s1-5779757/school-ai-education-students-teachers-poll-critical-thinking

Related reading: study mode vs. answer mode for kids, what to do when your kid used ChatGPT for homework, and setting an AI dose at home after PISA 2025.