Overview
- The Youth AI Safety Institute at Common Sense Media published its findings on Wednesday, July 15, 2026, after running roughly 2,500 tests on accounts configured for 11- and 15-year-olds with SafeSearch enabled.
- Researchers reported the AI Overview and AI Mode missed clear signs of suicidal ideation and psychosis, validated disordered eating, sometimes pointed to offline or unsafe resources, and in some cases supplied instructions that could enable abusive deepfakes.
- The institute found AI Mode completed homework across 180 academic prompts and AI Overview gave inconsistent factual answers, returning different responses 43 percent of the time and citing many unvetted social posts and forum content.
- Google formally disputed many of the findings, saying it could not reproduce several test outcomes, arguing the queries were narrow or ambiguous, and noting parents can block Search entirely on child accounts using existing controls.
- The report urges Google to add a clear toggle for parents and school administrators to disable AI Overview and AI Mode and warns regulators and educators to consider stronger safeguards because embedding adult-grade models by default exposes minors at scale.