Teach lateral reading on AI citations
Fact-checkers don't read a source harder. They leave it, open a new tab, and check who else says so. AI output deserves the same move.
Aug 23, 2026 · 1 min read
When Stanford researchers compared professional fact-checkers with smart students and even historians, the fact-checkers won by reading laterally: leaving the page almost immediately to see what the rest of the web said about it, instead of judging the page by its own polish (Wineburg & McGrew, 2019). AI output is the perfect training ground, because it's maximally polished and occasionally fabricates its references entirely.
In practice
- The rule for any AI-cited source: open a new tab and find it independently. If it can't be found, it doesn't exist for this essay.
- Run the drill once as a class: ask an AI for five sources on your current topic, then race to verify them. Finding the invented one teaches more than any warning.
- Teach the order: who published it, what do others say about them, then read the thing itself. Polish is not provenance.
Quick takeaway: judge a source by leaving it. Especially when a machine handed it to you.