A national survey found teen chatbot use for schoolwork doubled

The classroom note

A parent hearing that “a quarter of teens use ChatGPT for schoolwork” might picture a settled, universal habit. Pew Research Center's survey, published 15 January 2025, put a specific number on that impression: 26% of US teens aged 13 to 17 said they had used ChatGPT to help with schoolwork, against 13% asked the same question in 2023, based on 1,391 teens surveyed through Ipsos's KnowledgePanel between 18 September and 10 October 2024. That means 73% of teens still reported no such use, and the doubling describes a share that remains a minority.

What the evidence says

The survey is a self-report, not an observed measure of actual use, frequency or the quality of what students produced with it; Pew is explicit that the question asked whether a teen had ever used the tool for schoolwork, not how often or for what. The reported rate varied by group: 31% among Black and Hispanic teens compared with 22% among White teens, and 31% in grades 11 and 12 against 20% in grades 7 and 8, with no significant difference by gender or household income in this sample. Pew's own description of how it conducts panel-based surveys sets out address-based recruitment as the organisation's general standard for probability samples, distinct from the Ipsos-run teen panel this particular report drew on; the method still rests on parental consent and self-report by a minor, a different evidentiary weight than an RCT or an administrative record.

The implementation question

A use figure like this says nothing about what students are doing with the tool once they open it: checking a definition, drafting an outline, or having it write a paragraph outright are three different acts a single yes/no question cannot distinguish. A school reading this figure to justify a policy is really choosing between two different questions, how many students use this and what are they doing when they do, and the survey only answers the first.

What holds and what fails

The finding holds as a solid national estimate of self-reported use among teens with internet-connected households, at one point in time, drawn from an established panel method. It fails as a measure of learning impact, homework quality, or what any individual teacher is seeing in their own classroom, and the demographic gaps it reports describe correlations in this sample rather than caused differences. Reading “double” as “twice as much cheating” is an inference the survey itself does not support.

• Does this figure describe use, or does it describe an effect on learning or grades?
• Which students does the reported gap describe, and could access or awareness explain it as much as intent?
• Is this self-reported behaviour, or an observed or logged measure?

A use rate is a starting point for a conversation with students and families, not a conclusion about what that use is doing to them.