The share of American teenagers using artificial intelligence (AI) for homework rose from 40 per cent in May 2025 to 49 per cent in May 2026, and the increase has been uneven across income groups, according to a report released this month by RAND, a US-based nonprofit research organisation.

The report, Higher-Income Parents Use AI More and Encourage Their Children to Do the Same, by researchers Heather L Schwartz and Melissa Kay Diliberti, found that 57 per cent of teenagers from families earning US$150,000 or more a year reported using AI for homework in 2026, compared with 43 per cent of teenagers from families earning less than US$50,000. RAND described this as a 13-percentage-point gap. In the May 2025 survey, the researchers found no statistically significant difference between the two groups.

AI use for homework among teenagers in the highest-income group grew by 17 percentage points over the period.

Parents' AI use rising fastest at the top

The study draws on four nationally representative surveys of US parents and teenagers aged 12 to 18, conducted through the RAND American Parents Panel and the RAND American Youth Panel. The 2026 rounds covered 1,000 parents and about 1,800 teenagers.

The share of parents who reported using AI for work or personal reasons rose from 54 per cent in February 2025 to 72 per cent in May 2026. The rise was driven by parents who use AI at least once a week.

AI use grew among parents at all four income levels, but grew most among those in families earning US$150,000 or more. In 2026, 84 per cent of parents in that group used AI, compared with 63 per cent in the lowest-income group. The gap between the two groups widened from 15 percentage points in 2025 to 21 percentage points in 2026. According to the report, this widening came almost entirely from more high-income parents becoming frequent users.

The researchers stated that they presume higher AI use among higher-income parents stems from greater use of AI in white-collar, higher-paying jobs.

Who is encouraging children to use AI

In May 2026, parents were asked what they tell their children about using AI for school or homework. Overall, 56 per cent said they encourage it to some degree, 24 per cent said they discourage all AI use for schoolwork, and 20 per cent said they had not discussed it with their children.

The responses differed by income. About two-thirds (67 per cent) of parents in the highest-income group said they encourage their children to use AI for schoolwork, compared with 48 per cent of parents in the lowest-income group. Most parents who encouraged AI use said they did so with limits, such as using it for brainstorming or checking work rather than "to think for them". However, 17 per cent of the highest-income parents said they encourage their children to use AI wherever possible within school rules.

Parents' own AI habits were closely linked to what they told their children. Among parents who use AI frequently, 76 per cent encouraged their teenagers to use it for schoolwork, including 22 per cent who encouraged use wherever possible. Virtually no parents who never use AI encouraged use wherever possible.

When the researchers examined family income, parents' own AI use, parents' views on whether AI harms critical thinking and whether they see AI use as cheating together, only two factors predicted encouragement: how often parents used AI themselves and whether they considered AI use to be cheating. The authors call this a "do as I do" relationship.

What happens inside the household

The researchers were able to link 533 teenagers with 427 parents from the same households in 2026. RAND notes this matched sample was not designed to be nationally representative, was not weighted and is relatively small, and that its results should be read with caution.

Within these households, 63 per cent of teenagers whose parents were frequent AI users reported using AI for homework in at least some classes, compared with 40 per cent of teenagers whose parents never used AI.

Teenagers' perception of their parents' stance showed the strongest link. Of teenagers who saw their parents as encouraging AI use, 81 per cent used AI for homework, compared with 38 per cent of those who saw their parents as discouraging it. According to the report, teenagers' perceptions of parental attitudes were more strongly correlated with their own AI use than any other factor examined, including sex, family income and parents' views on critical thinking, with the exception of age.

What parents said and what teenagers heard often did not match. Among parents who said they encourage AI use wherever possible, only 25 per cent of their teenagers said their parents encouraged them. Another 27 per cent of those teenagers said their parents had discouraged AI use.

Mixed signals from school and home

Across the full sample, 78 per cent of teenagers said at least one type of person in their lives, such as parents, teachers, friends or school staff, had discouraged them from using AI for school, while 53 per cent said someone had encouraged them. Teachers were the most commonly cited discouragers (57 per cent), and friends the most common encouragers (36 per cent).

Only 33 per cent of teenagers who had no one encouraging them reported using AI for homework. When parents were among the encouragers, between 55 and 77 per cent of teenagers reported using it.

Wider gap, but uncertain consequences

RAND does not conclude that higher AI use among wealthier students will translate into an academic advantage. The report sets out two possibilities.

The authors point to emerging research, including four randomised controlled trials and one large quasi-experimental study, indicating that unrestricted AI use during a student's first attempt at a task, such as drafting an essay or solving a maths problem, harms performance. If that holds, they write, greater AI use among higher-income teenagers could lower their academic performance and narrow the long-standing income-based achievement gap.

Alternatively, the gap could widen if higher-income teenagers use AI in ways that do not outsource critical thinking, or use it to support college applications and career planning. The authors also cite smaller evidence that AI can support learning when it supplements a student's own work, for example by offering hints or asking questions.

The report recommends that states and school districts give schools specific talking points for students and families on why teenagers should not use unrestricted AI for first attempts at schoolwork. It also calls for policy controls on school-issued devices and for parental tools to set limits on AI use on personal devices.

"Perhaps the most important takeaway of this report is that parents have as vital a role as schools, if not more so, in conveying messages about responsible AI use to their teens," the authors write.

Other studies point in different directions

The RAND findings measure whether teenagers use AI for homework at all. A Pew Research Center survey of 1,458 US teenagers aged 13 to 17, conducted from September 25 to October 9, 2025 and published in February 2026, measured how heavily they rely on it and found a different income pattern. Fifty-four per cent of teenagers said they used chatbots for help with schoolwork. Twenty per cent of teenagers in households earning under US$30,000 said they do all or most of their schoolwork with chatbots' help, compared with 15 per cent in households earning US$30,000 to US$74,999 and 7 per cent in households earning US$75,000 or more.

A separate 2026 survey of 1,017 US teenagers by Common Sense Media's Youth AI Safety Institute found that 70 per cent use AI for schoolwork, and that 70 per cent said their teachers had never discussed how to use the technology safely.

The Indian picture

Comparable national data on AI use by household income among Indian school students is not yet available. Existing surveys indicate high exposure, alongside gaps in understanding and infrastructure.

The Bharat Survey for EdTech (BaSE) 2025 by the Central Square Foundation, conducted between August 2025 and January 2026 among 15,000 parents, children and teachers from low-income households across 10 states, found that 63 per cent of children use EdTech. Of these, 35 per cent use generative AI for learning, and 69 per cent of that group use it daily. Three-fourths of children who were aware of generative AI confused it with internet search. While 72 per cent of children had access to a smartphone at home, most access was shared.

A survey by the Centre for Policy Research and Governance (CPRG), released in February 2026, found that about half of private school students in Delhi use generative AI tools multiple times a week, mainly for information search and writing help.

School infrastructure remains uneven. According to UDISE+ 2024-25 data from the Ministry of Education, 57.9 per cent of India's 14.7 lakh schools have functional computers for teaching and 63.5 per cent have internet access. Government schools trail private unaided schools on both counts: 52.7 per cent against 72.5 per cent for functional computers, and 58.6 per cent against 77.1 per cent for internet. Across states, the share of schools with functional computers ranges from 99.1 per cent in Kerala to 23.8 per cent in Bihar.

About the study

The RAND report was sponsored by the Gates Foundation. The 2025 and 2026 youth samples differed slightly: the 2025 survey covered youth aged 12 to 21 and was restricted to those aged 12 to 18 in grades 5 to 12 for comparison, while the 2026 survey included oversamples from California, Florida, New York and Texas. The researchers did not test changes between the two years for statistical significance, and note that some differences over time may reflect methodological differences rather than real change. Family income data was collected after the surveys were conducted.