Published on 2026-07-25
How do professional fact-checkers actually check a claim?
Trained fact-checkers barely read the page in front of them. A Stanford study explains why leaving it is the smarter move.
Someone in your family group chat drops a claim that sounds too specific to be invented: a number, a name, a date. The instinct almost everyone follows next is to reread it, slower this time, hunting for the detail that will finally give it away. It rarely shows up. Words sitting on a screen don't confess anything, no matter how many times you scroll back over them.
Stanford researchers Sam Wineburg and Sarah McGrew ran an experiment in the late 2010s that explains why staring harder gets you nowhere. They sat down ten professional fact-checkers, ten PhD historians and twenty-five Stanford undergraduates, one at a time, and asked each of them to judge which of two websites on the same issue deserved more trust. The historians and the students did what most of us do: they stayed on the page, scrolling up and down, weighing the tone, the design, the confidence of the writing. Fifty percent of the historians and sixty-five percent of the students ended up trusting the less reliable of the two sites. The fact-checkers got it right, every single one of them, and did it within seconds. Their advantage wasn't sharper reading. It was that they barely read the page at all.
What tripped up the students and the historians wasn't unfamiliarity with the topic. It was the surface: a clean layout, a confident 'About' page, footnotes pointing to official-sounding institutions. That kind of polish costs very little to fake and says nothing about who's actually behind a page, yet our instincts treat it as a stand-in for credibility, because most of the time, in daily life, the two do line up. Online, that shortcut quietly breaks.
Leaving the page is the whole trick
What the fact-checkers did instead is what Wineburg and McGrew, writing up the results in Teachers College Record in 2019, called lateral reading. The moment they landed on an unfamiliar site, they opened new tabs and went looking for what independent sources said about it, before forming any opinion of their own. Not about the topic on the page. About the page itself: who runs it, who funds it, what people with no stake in the answer have to say about its record. The researchers found something almost paradoxical: the group who read the least on the original page ended up understanding the most about it.
The same move works on one claim, not just one site
You don't need a whole website in front of you to use this. A single claim responds to the exact same treatment, and it costs about as much effort. Pull out the specific, checkable part: a name, a place, a number, an event. Type it into a search, not into the reply box under the post where you found it. What comes back tells you most of what you need. If something genuinely significant happened, independent outlets, official sources or specialists in the field tend to have noticed it too, in their own words, around the same time. If the search turns up nothing but the original post and a stack of accounts repeating the identical sentence, that isn't three confirmations. It's one claim, echoed.
Picture a specific version of this: a message claims a factory upstream released something into the local water supply overnight. Search those exact details, the plant, the town, the night in question, and you'll typically find one of two things. Either regional news, a water authority or nearby residents are already discussing it independently, which suggests something real is unfolding. Or the trail runs cold right after the original post: no local coverage, no official statement, nothing from the people who live closest to a supposed disaster and would be first to notice. That kind of silence isn't proof that nothing happened. It's a strong reason to wait before you're the one who spreads it further.
There's a small discipline hidden in this that's worth naming: leaving the page means sitting, briefly, with not knowing. You've stopped reading the thing that convinced you and started reading what everyone who isn't invested in convincing you has to say instead. Sometimes that confirms your first instinct. Sometimes it demolishes it. Both outcomes count as the method working, because the point was never to catch the claim lying. It was to stop deciding based on how persuasive one single page happens to look.
A convincing page and a true one are not the same object, and only one of them survives a second opinion.
Next time something lands that you can't quite place, resist the sixth reread. Open a tab instead. Two minutes spent elsewhere will tell you more than twenty spent staring at the same four sentences ever will.
Sources
- Lateral Reading and the Nature of Expertise (Wineburg and McGrew, Stanford Digital Repository) (stacks.stanford.edu)
- Lateral Reading: Be a Pro Fact Checker (The Chicago School Library) (library.thechicagoschool.edu)