By Dorothy Lepkowska
Publication Date: 2026-09-17 23:01:00
Neither Google nor the rival AI engine, Perplexity, are safe for children according to tests carried out by the Youth AI Safety Institute.
A new risk assessment of Perplexity found that even when young people disclose their age, it largely treats them like adults and delivers unevenly on its core promise of more accurate answers.
It also repeatedly failed to provide useful help when testers simulated 15-year-olds in crisis. These problems prompted the Institute, a not-for-profit based in California, to assign Perplexity its lowest rating of Unacceptable Risk.
Geoffrey A Fowler, the Institute’s Head of Public Engagement, said Perplexity was tested over the same period as Google Search, running more than 1,200 search using the same, or functionally similar, prompts.
He wrote in a blog: “With reportedly 100 million monthly active users, Perplexity is much smaller than Google—but our goal was to test the field for approaches to AI search that could be better for kids than the AI Overview and AI Mode that Google has made the default for billions.
“Unlike Google Search, for instance, Perplexity offers an academic mode marketed to students. Perplexity’s strengths and weaknesses turned out to be different from Google’s, even though our rating is the same.“
What the Institute found:
Perplexity often detects kids in crisis but fails in its response. Across 114 crisis prompts, its default chat supplied a crisis line or medical referral 94% of the time—better than Google,…



