A national survey counts how many US teens use a chatbot

The design

On 9 December 2025, the Pew Research Center published 'Teens, Social Media and AI Chatbots 2025', by researchers Michelle Faverio and Olivia Sidoti. The report presents an online survey of 1,458 U.S. teens ages 13 to 17, fielded 25 September to 9 October 2025 through Ipsos, which recruited respondents via parents enrolled in its KnowledgePanel, a probability-based panel built from random address sampling. The survey was weighted to be representative of U.S. teens living with a parent, and the research was reviewed by an external institutional review board, Advarra.

What the evidence says

The report's own figures: 64% of teens say they have ever used an AI chatbot, and about three in ten say they use one daily. Among six named chatbots, ChatGPT was by far the most used at 59%, followed by Gemini (23%) and Meta AI (20%); the survey states 'fewer' teens use Copilot, Character.ai and Claude. On Character.ai specifically – the one companion-oriented product Pew measured by name – the report breaks use out by household income: 14% of teens in households earning under $75,000 report using it, roughly double the 7% among teens in households earning $75,000 or more. Pew's own text groups Character.ai alongside ChatGPT simply as 'chatbots' and does not ask teens why they use each one.

What it asks of people

A survey like this asks respondents only to recall and self-report use; it does not ask what a teen does with Character.ai once inside it, whether that use resembles a companion relationship or casual entertainment, or how long a session lasts. The income split Pew reports is descriptive, not explained by the survey itself, and should not be read as evidence about why lower-income teens report more use, since the questionnaire did not probe reasons.

Privacy and safeguards

Pew's report is a survey of attitudes and self-reported use; it does not evaluate any chatbot's data practices, age-verification method or safety design, and it does not ask teens' parents whether they were aware of the use reported. Its methodology page, current as retrieved on 16 September 2026, states the questions and full results are published separately alongside the report.

• Does grouping Character.ai with ChatGPT and Gemini as one category of 'chatbot' obscure how differently teens use each?
• What would the income gap in Character.ai use look like paired with a question about why teens chose it?
• How would self-reported daily use compare with what a platform's own usage logs would show?

The survey establishes a national, probability-based headline figure for teen chatbot use and a specific, attributable income split for one companion-oriented product, but it stops well short of explaining what that use involves or means to the teens reporting it, and it sits within Pew's broader ongoing programme on teens and technology.