Before your kid ever opens a chatbot, they're already using AI — every time they search. Here's the skill that matters now, and the one rule that teaches it: the Two-Source Rule.

Ask most parents when their child started using AI and they'll think of the day a chatbot app appeared on the phone. But for a lot of kids, AI arrived much earlier and much quieter — at the top of an ordinary search. In Common Sense Media's 2026 survey of children ages 9 to 17, AI-generated search summaries were the single most common type of AI kids encountered: 75% had used or interacted with them. That's more than chatbots, image generators, or anything else.

This matters for homework more than almost anywhere else. When your child looks up a date, a definition, or "why did the Roman Empire fall," an AI-written summary now often sits above the actual search results — a confident paragraph that answers the question so completely there's little reason to scroll, click, or read a source. For a kid doing research, that's a real shift: the skill used to be finding good sources; now the skill is checking the answer that's handed to you.

The good news is that the fix is simple, teachable, and works at every age. It's one habit — the Two-Source Rule — and this article is about how to teach it.

When an AI summary appears above search results, it's doing something different from a normal link list: it reads across web pages and writes a fresh answer in its own words, usually with a few links alongside. Google describes its AI Overviews as an AI-generated snapshot with links — and, in its own help pages, says plainly that AI responses may include mistakes. That's the company that builds it telling you the answer isn't guaranteed.

The lesson for a student isn't "AI search is bad." It's "an AI summary is a starting point that can be wrong, so verify what matters."

The subtler problem isn't that AI summaries are sometimes wrong — it's that they're so convenient people never check. In an analysis of U.S. adults' Google browsing behavior, Pew Research Center found that when people saw an AI summary they rarely clicked a link — in that analysis just 1% clicked a link within the summary — and were more likely to end their browsing session right there. That study looked at adults, not children — but if adults stop at the summary, it's easy to imagine students doing the same during homework. The summary becomes the destination, not the doorway.

For a student, that's the habit to worry about. A kid who stops reading past the AI paragraph doesn't just risk copying an occasional error — they lose the muscle of checking a claim against a real source, which is the core of research and of thinking critically about anything online.

The rule is simple, and it matches Google's own advice for using AI answers: always check important information in more than one place. Turned into a family habit:

"If it matters" is the key phrase. Nobody needs to fact-check a movie's runtime. The rule kicks in for schoolwork, health, news, and anything a kid is about to write down, cite, or believe.

Here's the rule applied to an actual assignment question — "Why did the Roman Empire fall?":

Not all second sources are equal — a rule that just means "find any two things" is easy to game. A second source should be something your child actually opens and reads, and ideally it should be independent and credible — a published article from a reputable news organization, a government or public-health agency, an academic or scientific institution, or an established reference source — not just another AI-generated summary repeating the same claim. The point isn't to collect two links; it's to see the same fact stand up in two real, trustworthy places.

Research note: two independent, non-peer-reviewed academic audits give a sense of the failure modes (worth knowing, but early audits of specific question types, not proof about all AI search). One preprint auditing AI Overviews on baby-care and pregnancy questions found the answers were inconsistent with each other in 33% of cases; another 2026 preprint reported about 11% of statements were unsupported by the cited pages.

The rule is the same. How you teach it grows with them.

Ages 7–10. This is the age to plant the idea that the top answer is a "robot's best guess," not automatically true. Do it together: read the AI summary aloud, then click a real link and compare. Make finding a second source a small game rather than a chore, and keep research in a common area so you can model the click.

Ages 11–13. Homework gets real, and so does the temptation to paste the summary and move on. Name the Two-Source Rule explicitly and tie it to schoolwork: a summary can be a starting point for notes, but a citation has to come from a source they actually opened. This is the age to show them, with a concrete "these two answers disagreed" example, that AI summaries can be confidently wrong.

Ages 14–18. Teens can handle the why: AI writes a plausible answer by predicting text, which is how it can state something unsupported with total confidence. Push past "find two sources" to evaluating them — is the second source independent, or just repeating the same claim? This is media literacy, and it transfers far beyond homework into how they read the news and everything else online.

📎 Free download: "The Two-Source Rule" — a one-page guide with the rule, the worked example, what counts as a real second source, and the by-age steps.

Your child's first and most frequent AI tool probably isn't a chatbot they chose — it's the summary that appears whether they asked for it or not. You may not always be able to control when AI summaries appear, either — they show up across different devices, browsers, school accounts, and search products, and the settings change often. So you don't rely on turning it off. You hand them one durable habit instead: the top answer is a starting point, and if it matters, you check it in a second place. Two sources. Teach that once, reinforce it as they grow, and you've given them a skill that outlasts any particular app.

This article is general guidance and is not legal, medical, or mental-health advice. Disclosure: AI by Age is run by people who work in AI and build AI-related products. We take no AI-vendor sponsorships or affiliate payments. Full disclosure on our About page.