The AI Homework Gap: Lower-Income Teens Lean on AI 3x More

One in five teens in US households earning under $30,000 a year say they do all or most of their schoolwork with an AI chatbot’s help. In households earning $75,000 or more, it’s 7%. That three-to-one ratio, from Pew Research Center’s February 2026 report, was quoted by Chelsea Clinton at Common Sense Media’s Kids and AI Summit on October 28.

Here is the detail most retellings missed. Teens in middle-income homes, $30,000 to $75,000, were at about the same rate as the lowest-income group. The AI homework gap isn’t between poor kids and everyone else. It’s between affluent kids and most kids. This article looks at why reliance may concentrate where support is thinnest, why that matters for learning, and what schools and families can do.

Key Takeaways

• Pew: 20% of teens in households under $30,000, a similar share at $30,000 to $75,000, and 7% at $75,000+ do all or most schoolwork with chatbots.
• Black and Hispanic teens are more likely than White teens to use chatbots for schoolwork and to call them very helpful.
• A June 2026 working paper on 26,811 Chinese students found AI adoption raised homework scores 18% but cut exam scores 20%, with losses concentrated among students who appeared to outsource.
• A World Bank pilot in Nigeria found teacher-guided AI tutoring produced gains of about 0.3 standard deviations in six weeks.
• The difference is design: AI that does the work widens gaps; AI that explains under guidance may narrow them.

What the AI Homework Gap Is

The AI homework gap is the difference, by family income, in how heavily teens rely on AI chatbots to complete schoolwork. Pew’s demographic analysis of 1,458 teens surveyed in fall 2025 put numbers on it:

• One in five teens in households under $30,000 do all or most of their schoolwork with AI chatbots’ help.
• A similar share in households from $30,000 to just under $75,000 say the same.
• In households at $75,000 or more, 7% do.

Across all teens, Pew’s main report found one in ten do all or most schoolwork with chatbots, 21% some and 23% a little; 45% haven’t used them this way.

Race and ethnicity show a related pattern. About six in ten Black or Hispanic teens have used chatbots for schoolwork help, against roughly half of White teens. Black and Hispanic teens were also more likely to say chatbots have been extremely or very helpful for schoolwork, and more likely to do most of their schoolwork with them.

TIME’s coverage of the summit reported Clinton framing this as an equity issue. The numbers support the framing.

Why Reliance May Concentrate Where Support Is Thinnest

Pew’s survey measures who relies on AI, not why. Several explanations fit the data, and other surveys add pieces.

AI is the help that’s available. A teen without a parent who can explain algebra, a paid tutor or an easy path to a teacher’s office hours has one tool that answers at 11 p.m. for free. In that light, heavy use can be resourcefulness, not laziness.

Kids who struggle use it more. Common Sense Media’s June 2026 census found 56% of kids who have trouble focusing on school assignments use AI for schoolwork weekly, against 45% of kids who don’t. Similar gaps appeared for kids who find math or essay writing hard. Interestingly, that census found overall AI use similar across income levels; the difference Pew captures is in how heavily it’s used for schoolwork.

Fewer conversations at home. Pew’s parent survey found parents in households under $30,000 were less likely than those at $75,000 or more to have talked with their teen about chatbots (43% vs 56%). Less guidance means fewer agreed limits on what AI is for.

These are plausible mechanisms, not proven ones. They point in the same direction, though: the students with the least human support are the ones most likely to hand the work to a machine.

Why It Matters: The Learning Penalty

If AI-completed homework produced the same learning as self-completed homework, the gap wouldn’t matter much. The best real-world evidence says it doesn’t.

A June 2026 CEPR discussion paper by David Strömberg, Victor Lei and Yanhui Wu followed 26,811 Chinese students in grades 7 to 12 over 30 months. AI adoption raised homework scores by 18% and cut completion time by 30%, but lowered monthly exam scores by 20% within six months. High-stakes entrance-exam scores fell 18% and 24%, with the full penalty emerging after about two years. Losses were concentrated among roughly 80% of AI users whose fast, high-scoring homework suggested outsourcing. Students who kept homework time similar to non-users showed only small losses.

The study is not peer-reviewed, comes from one country’s school system, and isn’t randomized. But it’s large and long, which is rare in this research. Our breakdown of the 18% homework, 20% exam study covers it in depth.

Put it together with Pew and the risk is clear: if lower- and middle-income teens outsource more, they may pay more of the exam penalty, months later, when it’s hard to trace back.

The Other Side: Guided AI Can Narrow Gaps

The same technology can help when the design is different. In a World Bank pilot in Edo State, Nigeria, secondary students were randomly assigned to a six-week after-school program using a GPT-4-based tutor with teacher support. The researchers reported learning gains of about 0.3 standard deviations, which they describe as roughly two years of typical learning, outperforming about 80% of education interventions in their comparison database. Girls, who started behind, appeared to gain more, and each extra day of attendance was linked to larger gains.

That evaluation is short, from one setting, and was reported before full publication. Its lesson is about structure: teachers framed the prompts, students engaged with explanations, and the AI didn’t hand over answers.

Outsourcing Versus Tutoring: The Decision Matrix

Pattern	What the student does	What the evidence suggests	Equity effect
Outsourcing	Pastes the assignment, submits the answer	Homework up, exam learning down (CEPR, 2026)	Widens gaps if lower-income teens rely more
Checking	Does the work, asks AI to check or explain errors	Keeps practice; plausible benefit, less studied	Neutral to positive
Guided tutoring	Works through problems with AI prompts set by a teacher	~0.3 SD gains in a six-week pilot (World Bank, 2025)	Can narrow gaps
No access	No AI, no other help	Status quo	Gaps persist

What Schools Can Do

Assess learning in class. If homework can be outsourced, grades built on homework stop measuring learning. More in-class work, oral checks and drafts shift weight to what students can do.

Teach the tutoring pattern explicitly. Show students prompts that make AI explain rather than answer, and practice them in class, especially for students without help at home.

Offer human help where AI fills the gap now. Homework clubs, peer tutoring and teacher office hours give students who rely on AI an alternative. The American School Counselor Association reports 372 students per counselor nationally in 2024–25, so school staff are already stretched; targeting matters.

What Families Can Do

Watch time, not just grades

The CEPR finding suggests a useful household signal: homework that suddenly takes far less time while scores stay high. Ask your teen to walk you through how they did one assignment.

Set the “explain, don’t answer” rule

Agree that AI can explain a concept, check work or quiz, but the answer is your teen’s. Our guide to when AI homework tools help versus hurt has example prompts.

Use AI as the tutor you can’t afford, on purpose

If paid tutoring isn’t in reach, a well-prompted chatbot can explain a step at 11 p.m. The point is to use it the Nigeria way, not the outsourcing way.

What not to do

Don’t treat heavy AI use as cheating by default. For many teens it’s the only help available. Banning it without offering another source of help may simply leave them stuck.

What to Watch For Over the Next 3 Months

• Week 4: Your teen can show you one assignment where AI explained something and they wrote the answer themselves.
• Month 2 red flags: Homework time drops sharply while grades hold, then a test or quiz score falls.
• Month 3 self-check: Do you know what your teen’s school does to check learning in class? If not, ask at the next conference. For Pew’s full schoolwork numbers, see 1 in 10 teens do most schoolwork with AI.

Frequently Asked Questions

Do low-income teens use AI more for homework?

Yes, according to Pew’s February 2026 report. One in five teens in households under $30,000 do all or most schoolwork with chatbots, as do a similar share at $30,000 to $75,000, versus 7% in households at $75,000 or more. The data are self-reported by US teens 13 to 17.

Is using AI for homework bad for learning?

It depends on how. A June 2026 study of 26,811 Chinese students found AI raised homework scores but lowered exam scores, mainly among students who appeared to outsource. Students who kept spending similar time on homework showed only small losses. Guided tutoring showed gains in a Nigerian pilot.

Why do lower-income teens rely on AI more?

Pew didn’t ask why. Plausible reasons include less access to tutors or help at home, fewer family conversations about AI (43% vs 56% of parents by income), and greater struggle with schoolwork, which other surveys link to heavier AI use. These are explanations, not proven causes.

Can AI help close achievement gaps?

Possibly, if it’s used to explain rather than answer. A World Bank pilot in Nigeria using teacher-guided AI tutoring found gains of about 0.3 standard deviations in six weeks. Unstructured use that completes assignments points the other way. Design and supervision decide which happens.

What should parents do if their teen does most homework with AI?

Ask them to walk you through an assignment, agree that AI explains but the answer is theirs, and watch for homework time dropping sharply. If they lack other help, look for school homework clubs or tutoring rather than banning the tool outright.

---

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. Pew Research Center. (2026, February 24). “Demographic differences in how teens use and view AI.” Pew Research Center. https://www.pewresearch.org/internet/2026/02/24/demographic-differences-in-how-teens-use-and-view-ai/
2. McClain, C., Anderson, M., Sidoti, O., & Bishop, W. (2026, February 24). “How Teens Use and View AI.” Pew Research Center. https://www.pewresearch.org/internet/2026/02/24/how-teens-use-and-view-ai/
3. Pew Research Center. (2026, February 24). “What parents say about their teen’s AI use.” Pew Research Center. https://www.pewresearch.org/internet/2026/02/24/what-parents-say-about-their-teens-ai-use/
4. Luscombe, B. (2026, October 30). “‘Social Media on Steroids.’ Advocates for Children Call for Guardrails Around AI.” TIME. https://time.com/article/2026/10/30/ai-kids-common-sense-media/
5. Common Sense Media. (2026). “The Common Sense Media Census: AI Use by Tweens and Teens, 2026.” Common Sense Media. https://www.commonsensemedia.org/sites/default/files/research/report/2026-ai-use-by-tweens-and-teens-1.pdf
6. Strömberg, D., Lei, V., & Wu, Y. (2026, June). “The Generative AI Learning Penalty: Evidence from Chinese Secondary Education.” CEPR Discussion Paper 21577 (not peer-reviewed). https://econpapers.repec.org/RePEc:cpr:ceprdp:21577
7. De Simone, M. E., Tiberti, F., Mosuro, W., Manolio, F., Barron, M., & Dikoru, E. (2025, January 9). “From chalkboards to chatbots: Transforming learning in Nigeria, one prompt at a time.” World Bank Blogs. https://blogs.worldbank.org/en/education/From-chalkboards-to-chatbots-Transforming-learning-in-Nigeria
8. American School Counselor Association. (2026). “School Counselor Roles & Ratios.” ASCA. https://www.schoolcounselor.org/About-School-Counseling/School-Counselor-Roles-Ratios