Public Opinion Survey on Parents’ Views on Children and Teen’s Use of AI

KEY TAKEAWAYS:

• Only 16% of parents feel certain they can keep their kids safe online, while nearly half of parents (49%) are worried they can’t keep their children safe online.
• 6 in 10 (61%) view AI as bad for kids, and only 20% say it is good for children and teens.
• Only 1 in 10 Americans give tech companies as core of 8 or above on a 0 to 10 scale, where 10 means they completely trust the tech companies that create apps, websites, and content that kids use to prioritize child safety, and 0 means they do not trust the companies at all todo this. Three-quarters gave responses of 0 to 5.
• 45% of parents say kids getting inappropriate romantic or sexual responses from AI chatbots is a widespread problem
  
  • 42% of parents say the same of AI chatbots not reporting troubling behavior from kids, including talk of suicide
  • 42% of parents say kids developing intense emotional bonds with AI chatbots is widespread or happening here and there.
  • Only 5% or fewer parents say each of these interactions are only rarely or not at all problems.
• Democrats are more likely than Republicans to think they can keep their kids safe online, but partisans feel similarly about the risks of AI for kids and share the same lack of trust in companies to prioritize child safety.
• On changing Section 230: Nearly 3 in 4 voters (73%), including 81% of Democrats and 71% of Republicans, support holding social media companies accountable for various situations where users may be harmed by content or exposed to illegal or prohibited content.
• Eight in 10 support holding social media companies accountable when the harm involves kids under age 13 or allowing children and predators to interact.

Frequency Questionnaire (October 22-27, 2025)

‍

‍

For each situation below, please indicate if you support or oppose holding social media companies legally accountable.

‍

‍

‍

For each of the following, please say whether you think it is a widespread problem, a problem here and there, very rarely a problem, or not a problem at all for kids on the internet.

‍

‍

‍

‍

Survey methodology: GQR and American Viewpoint conducted a 20-minute mixed-mode text-to-web and online survey among 800 registered voters in theUnited States from October 22-27, 2025. The survey used voter file sample to reach text-to-web respondents and contacted online respondents from a panel sample of United States residents, who were then matched to the voter file to verify their eligibility. There are 332 parents of children under age 18, 296parents of children between ages 5 and 18, and 239 parents of children between ages 11 and 18.

Data are weighted on the demographics of registered voters, including gender, age, race, education, and parental status, and the non probability online panel data are additionally calibrated to the probability-based text-to-web data based on frequency of survey participation and political interest.

 If this entire survey was based on a probability sample, the margin of error would be+/- 4.5 percentage points for the full sample and +/- 7 percentage points for the parents subsample, The margin of error is higher among subgroups. Because the online sample is based on those who initially self-selected for participation in the panel rather than a probability sample, no estimates of sampling error can be calculated.

In this report, results are expressed as percentages unless otherwise noted. Results may not add to 100% due to rounding or multiple responses. Net results cited in the text may not exactly match individual results shown in the charts due to rounding.

‍