Report: Google's AI Search Poses Unacceptable Risk to Children

Google's built-in AI search features, AI Overview and AI Mode, pose an unacceptable risk to children and teenagers, according to a report published on July 15, 2026 by Common Sense Media. It is one of the most detailed independent safety assessments yet of AI tools that cannot be turned off in Google Search, even though three-quarters of American teens already use them.

Common Sense Media, a US nonprofit focused on children's safety in digital media, has spent months publishing assessments of popular AI chatbots and assistants. This time the focus was not a standalone app but a feature built directly into a search engine used by billions of people every day.

What the tests found

Researchers examined how AI Overview and AI Mode handled questions about mental health, addiction, violence and schoolwork. The results were unambiguous: both features failed all five of the most severe risk categories and fell short on seven of the eight general principles for responsible AI behavior toward children that Common Sense Media applies when evaluating any product.

The most serious concerns involve recognizing mental health crises. The algorithms missed nearly a third of direct mentions of suicidal thoughts and half of the indirect signals, and in one test described a symptom of an eating disorder as normal. On top of that, the system repeatedly directed users to a crisis hotline number that stopped working in 2023.

Doing homework and spreading false information

The second major problem is the impact on learning. AI Mode, available on laptops schools issue to students among other devices, completed all 180 tested math assignments and essays for them instead of helping them work through the problems themselves. Researchers note that this kind of feature undermines the whole point of homework as a learning tool.

Testers also found that answers to identical questions could vary significantly between attempts, and that the system pulled information from online forums and social media without any editorial verification. In several cases the AI provided instructions useful for creating deepfakes, which can be used for peer cyberbullying.

No way to turn it off

A key point of criticism is that AI Overview appears automatically above search results and cannot be disabled, even with SafeSearch, the mode meant to protect younger users, turned on. Parents and schools therefore have no real control over whether a child encounters an AI-generated summary at all.

It's deeply concerning how poorly these widely available tools perform - Justin Reich, MIT

Nobody was asked, nobody got to click a button saying 'do you want AI Overview' - Justin Reich, MIT

Robbie Torney of Common Sense Media told media outlets that the feature was rolled out by default to all search users, with no exception for child accounts. Google responded that it considered the test queries narrow and artificially constructed, and said the company was unable to reproduce some of the responses described in the report.

What it means for Poland

The report focuses directly on the US market, but AI Overview and AI Mode run in the same form, impossible to switch off, in the Polish version of Google Search as well. Earlier research by Gemius found that one in three Polish children aged 7-14 already uses ChatGPT, and Poland has not yet joined the UN coalition for protecting children from AI, which makes the question of search engine makers' accountability just as relevant on the Vistula.

The EU's AI Act imposes obligations on providers of high-risk systems related to protecting vulnerable groups, including children, but search engines with built-in AI summaries are not clearly classified in that category. The Common Sense Media report could become an argument in the debate over whether regulation should extend to this kind of search feature and not just standalone chatbots.

For parents and schools in Poland, the practical takeaway is simple: SafeSearch and parental controls in their current form do not protect against content generated by AI Overview, because that layer operates independently of standard search filters. Experts cited in the report recommend that educational institutions treat AI Mode on school devices as a separate risk requiring additional settings beyond standard SafeSearch.

Google said it would review the report's findings but has not announced any specific changes to how AI Overview works, nor a timeline for adding an option to turn it off. Common Sense Media plans further rounds of testing as both features evolve, meaning the issue is likely to resurface with future search updates.