Students do not need to be told misinformation exists. They need a procedure they can run in thirty seconds, and one worked example of it failing.
The two methods with the best evidence behind them are lateral reading, where you leave the page to check who wrote it, and SIFT: Stop, Investigate the source, Find better coverage, Trace claims to the original. Both can be taught in a single lesson with nothing but a browser, and both work without any tool.
A class that learns to run a tool has learned nothing transferable, because the tool will change. A class that learns lateral reading keeps it for life. So the honest order is method first, and any software second, if at all.
Lateral reading means leaving the page you are judging. Instead of scrutinising the site's design and its About page, you open new tabs and find out what independent sources say about it. Professional fact-checkers do this reflexively, and studies of expert versus novice behaviour keep finding it to be the single biggest difference.
SIFT gives students four moves in an order they can remember: Stop before sharing, Investigate who is behind it, Find better coverage of the same claim, Trace quotes, numbers and images back to the original context. It is deliberately short, because a checklist longer than four items does not survive a real feed.
This matters in schools specifically, because detectors are being pointed at pupils. We measured how well AI-text detection works before deciding whether to ship it, and published every number.
The design was built to be fair to the human side: 120 comments posted before 30 November 2022, the day ChatGPT was released, so they cannot have been machine-written, drawn at random rather than hand-picked. Against them, 120 generated texts in three stances, including one instructed explicitly to sound like a human, matched across three length bands. Detection was not reliable enough to put in front of a user, so we did not ship it.
The consequence for a classroom is not subtle. Detectors are documented to falsely accuse writers who use English as a second language. A false accusation of cheating is not a smaller error than a missed case of cheating: it is the same error pointed at a student who did nothing wrong, with a disciplinary process attached. If your school uses such a tool, the raw data is here and it is usable in that conversation.
Useful: as the second step, after students have formed a judgment themselves. Run a claim through a verification tool and compare its evidence chain with what the class found. The interesting moment is a disagreement, because it forces both sides to show their sources.
Harmful: as the first step. A class that starts with a score learns to outsource judgment, which is the exact habit misinformation exploits. A truth score is a statement about evidence, not a verdict about truth, and no tool rules on intent.
What we do not have, said plainly: there is no school licence, no classroom management, no teacher dashboard and no age-verification layer. The free tier is 5 checks a day per user, which covers an exercise but not a term. If a school needs something structured, write to us and we will say honestly whether it exists yet.
Free: 5 checks a day, no account. Run one claim, look at the evidence chain, and see where it says a statement has no source behind it.
wyper