A youth-safety research body rated the artificial intelligence (AI) features built into Google Search an “unacceptable risk” for children, the lowest score on its scale, and found the tools also generated instructions for building deepfakes usable in harassment.

The Youth AI Safety Institute, part of nonprofit Common Sense Media, officially released the assessment on July 15, 2026. Researchers ran more than 2,600 searches over six weeks using accounts configured with SafeSearch for simulated 11 and 15-year-olds, and audited over 2,100 cited sources.

Google Search’s AI Overview and AI Mode failed all five of the report’s severe-harm “Red Lines” and scored poorly on seven of eight safety principles, the Institute said. Testers documented the tools missing explicit suicidal ideation, validating disordered eating, and treating forum posts as equal to medical institutions.

Both features produced instructions for creating deepfakes that could be used for bullying, according to the report. Synthetic-media generation is a recognized information-warfare and fraud vector, and the finding places a mass-market consumer tool alongside capabilities defense analysts track as adversarial threats.

Robbie Torney, who leads AI assessments at the Institute, said the group recommends children and teens not use the product. Unlike Google’s standalone Gemini chatbot, rated “high risk,” the search features cannot be disabled.

A Google spokesperson said the tests used ambiguous and contrived queries that do not reflect real use, and stated the company could not reproduce many results. Google said its Family Link controls can block Search access and that crisis hotlines appear when relevant.