Three Out of Four Kids Get Their Google Answers From AI. Here Is the One Habit Worth Teaching Before They Stop Checking.

The last time my youngest Googled something for a school project, she read the first thing on the screen and moved on. She did not click the link below it. She did not look for a source. She read the AI-generated paragraph at the top of the results and accepted it as the answer. I watched it happen in about twelve seconds. I did not say anything, because I did not fully understand what I was watching. A new report from Common Sense Media changed that.

Three out of four American kids ages 9 to 17 are now reading AI-generated answers every time they do a Google search, and most of them have no idea this is happening.

When your kid types a question into Google, the first thing they see is often not a link to a website. It is a paragraph written by Google’s artificial intelligence. This feature is called AI Overview. Google also offers AI Mode, which generates longer responses to more complex questions. Both features are on by default. Neither one asks kids, or parents, whether they want it. Your kid searches. AI answers.

In July 2026, Common Sense Media’s Youth AI Safety Institute published a risk assessment of both features after running more than 2,600 test interactions.^[1][2] The researchers submitted questions and scenarios the way a child would, testing whether the AI could recognize when a young person was in trouble, whether it encouraged academic shortcuts, and whether its answers were accurate enough to be trusted. They graded each tool against eight child safety principles and five definitions of severe harm.

The results landed badly. Both AI Overview and AI Mode violated seven out of eight child safety principles, and both failed all five of the institute’s definitions of severe harm.^[1] Justin Reich of MIT’s Teaching Systems Lab, cited in the report, called the findings “deeply disturbing.” I read that phrase first in the PBS NewsHour headline and assumed it was editorial language. After reading the actual results, I stopped assuming.

The specific failures the AI tools showed in testing matter because three-quarters of American kids are already using them, without choosing to, and without knowing what they are getting.

The most direct finding involves homework. The researchers submitted 180 math and essay assignments to the AI tools, the kind a kid might actually turn in for school. AI Mode completed all 180 of them.^[1] That is not a minor edge case. It means the AI is not summarizing or assisting. It is doing the work from start to finish. A child can take that completed assignment, submit it, and never write a sentence or solve a single problem themselves.

The safety failures are harder to sit with. AI Overview missed 29 percent of explicit statements about suicide and half of all indirect ones.^[1] These were not obscure signals the researchers had to hunt for. They were straightforward moments where a young person might be reaching out, and the AI responded as though nothing was wrong, or gave a generic answer that did not fit the situation. Both tools also gave responses researchers described as reinforcing harmful thinking around eating and other serious concerns.

The part that stays with me is the baseline stat: three out of four kids already use these features by default.^[1] There is no decision being made. They search, and AI answers, in the same spot where answers have always appeared. My youngest does not separate “the search engine is showing me this” from “someone verified this.” She has not been taught to. And the report makes clear the AI answering her was not built with a child’s judgment gaps in mind.

The skill is Source Checking: looking past the AI-generated answer to find and read the original source it claims to draw from, or to find a source the AI did not provide at all.

This sounds simple. It is not automatic. My middle daughter trusts what appears on the screen. She has not yet learned to separate “the search engine showed me this” from “this has been verified.” That is not a failure on her part. It is a habit she has not been taught yet. That is the opening for me as her parent.

Source Checking has three parts. First, recognizing that what appears at the top of a Google result might be generated by AI, not pulled from a specific expert or news organization. Second, finding where the AI says it got its information, when it says anything at all. Third, clicking through to that source, reading even one paragraph, and asking: does this actually say what the AI claimed? None of those steps takes more than a minute. Together, they are the difference between a reader who accepts a result and one who thinks about it.

More than 4 in 10 kids say no parent or guardian has ever talked with them about AI safety.^[3] That number stands out to me because it is not about kids misusing AI, or schools failing to teach it. It is about whether the basic conversation is happening at home at all. Source Checking is a small, concrete version of that conversation. It does not require a family meeting or a rule about screens. It is one question, repeated enough to become reflex: where did that come from?

A child building the Source Checking habit treats AI answers as starting points, not endpoints, and begins asking questions about what they are reading before they accept it or share it.

If your kid reads something and says “but where did that come from?” or “is that actually true?”, they are already practicing an early version of this skill. You will also notice it when they click a link inside an AI result to check the original article, or when they read a summary and say the headline does not quite match. These are small moments, but they reflect something real: your kid is learning to participate in information rather than just receive it.

A less obvious sign is frustration. When my oldest complains that two searches gave her different answers on the same topic, I used to treat that as impatience. Now I see it differently. Noticing a contradiction is a first step toward Source Checking. She is registering that the AI is not always consistent, which means she is not fully trusting it. That frustration, redirected, becomes: “They said different things, so let me find the actual source.” The frustration is not the problem. It is the beginning of the habit.

Pick one moment this week to show your kid what an AI search result looks like and walk through a source check together.

Show the label. Open Google with your kid and search for something they care about, not something school-related. Point to the AI-generated block at the top and name it: “That paragraph was written by an AI, not a person at a news organization or a research group.” Many kids have never heard this said out loud. They assume a search result has to come from somewhere specific. Naming the source, or the lack of one, changes how they see it, without making them afraid of it.

Click through once together. After you point out the AI answer, scroll down and click one of the links below it. Read one paragraph from the actual article together. Then ask: “Does this say what the AI said?” You do not need to make it a lesson. One demonstration, done casually in an ordinary moment, builds more than a ten-minute talk you prepare and never quite have.

Make one question the family default. Any time AI comes up at home, a search result, a chatbot answer, something your kid says they read online, ask: “Where did it get that?” You do not need to investigate every time. You are planting a question. Over weeks, your kid starts asking it themselves, before you do. That is when the habit has taken root.

Talking about AI with your kid does not have to start with a warning. It can start with curiosity. “Did you notice what that search result was?” is an easier opening than “I need to talk to you about something.” The Common Sense Media report is a reason to have a specific conversation. The specific conversation is about one habit: looking at what the AI gave you and asking where it actually came from.

I started paying more attention to how my girls use Google after reading this report. What I noticed surprised me less than I expected. They are fast. They find answers quickly, which means they accept answers quickly. The problem is not that they are lazy. The problem is that nothing in the experience of doing a Google search teaches them to slow down. And the AI features, which appear at the very top of the results page, make the answers look more authoritative than anything that came before them.

The skill I am working on with mine is not suspicion. It is one extra step. Where did that come from? Click through. Read one paragraph. Decide whether the source makes sense for the question they asked. That is Source Checking. It takes thirty seconds. But it is the difference between a kid who reads information and one who thinks about it.