You're previewing the Bennett demo family. Numbers, schedule, and kids are sample data.Start your own
Loading page…
Researchers gave the same website to professional fact-checkers, to historians with doctorates, and to Stanford undergraduates. The fact-checkers were faster and more accurate — and the first thing they did was leave the page.
You have probably been taught a checklist for deciding whether a website can be trusted. Something like: look at the design, read the About page, check whether it ends in .org, look for a padlock in the address bar, see whether it lists its sources.
Every item on that list has the same flaw, and once you see it you cannot unsee it.
All of them are controlled by whoever made the page.
A professional-looking design costs a few hours. Anyone in the world can register a .org — it has never been restricted to anything. The padlock means your connection to the site is encrypted, so nobody can read it in transit; it says precisely nothing about whether the words are true. And the About page is a reference letter that the person being checked has written about themselves.
So the checklist has you standing on the page, using materials supplied by the page, to decide whether to believe the page.
There is a better move, and it is not more careful. It is in a different direction entirely.
Researchers at Stanford ran a comparison. They took three groups — ten professional fact-checkers from major news organisations, ten historians with doctorates, and twenty-five Stanford undergraduates — sat each person in front of live websites, and recorded them working out whether the sites were trustworthy.
The historians and the students did what you were taught. They stayed on the site. They studied its look, read its About page, weighed up its domain name, scrolled its articles. Some of them were taken in — including by a site that turned out to be run by a public-relations firm on behalf of an industry.
The fact-checkers barely read the site at all. Within seconds of arriving, they left. They opened new tabs and searched for the organisation's own name, to see what everyone else on the internet said about it. Then they came back — already knowing who they were reading.
The name for it is lateral reading: instead of going deeper into a page, you go sideways off it, and let sources the page does not control tell you who wrote it.
Here are seven things people do when checking a source. The only question that matters about each one is whether it keeps you inside the page's own account of itself. Sort them and read what each is really worth.
Tap an item, then tap where it belongs
A site is beautifully designed, ends in .org, shows the padlock, and has an About page describing a team of eleven researchers with photographs. How much have you learned about whether to trust it?
An invented example, so nobody's real argument is on trial here. A page from the Institute for Everyday Nutrition says a common food additive was 'shown to triple' a health risk. Watch what four minutes off the page gets you.
The About page has ten paragraphs of detail about who runs the organisation. A web search of its name gives you three short results. Why is the search worth more, when it plainly contains less?
When the fact-checkers ran their searches, they did something else the other two groups did not. They did not click the first result. They read the whole results page first — the little descriptions, the site names, how the results disagreed with each other — and then chose where to go.
The researchers call this click restraint, and it takes about ten seconds.
It matters because the first result for an organisation's name is very often the organisation itself, which is the one page you have already decided is not evidence. What you want is further down: the news write-up, the encyclopaedia entry with its references, the other group saying the opposite. The interesting results are almost never at the top, because the top is where the most popular answer lives, and popularity is not what you came for.
Everything so far is easy to do on a page making a claim you dislike. You were already suspicious; the tabs practically open themselves.
The hard case is a page that tells you something you were hoping was true. A study that says the thing you already believed. A story that makes the person you already disagreed with look bad. In that situation you do not feel any urge to check, because nothing feels wrong — and nothing feeling wrong is not information about the page. It is information about you.
This is the difference between a skill and a weapon. Used only on the side you dislike, lateral reading will make you more confident and less correct, every single time, because you will have carefully checked one half of what you read and waved the other half through.
The rule that keeps it a skill is embarrassingly simple: run the check when you agree. If that feels like a waste of four minutes, that is precisely the reading on the instrument that tells you it isn't.
You search an organisation's name. The first result is its own website; results two to five are other people writing about it. Where do you go, and why?
You’re previewing as a parent — nothing here is recorded.
Wineburg, S., & McGrew, S. 'Lateral Reading and the Nature of Expertise: Reading Less and Learning More When Evaluating Digital Information.' Teachers College Record, 121(11), 2019 (first circulated as a Stanford History Education Group working paper in 2017). Ten professional fact-checkers, ten historians holding doctorates and twenty-five Stanford undergraduates were observed evaluating live websites. Fact-checkers reached accurate conclusions faster; they read laterally, leaving a site quickly to consult other sources about it, while historians and students read vertically, staying on the site and judging it by its own contents. One of the sites used in the study was operated on behalf of an industry by a public-relations firm. 'Click restraint' — scanning the full results page before clicking rather than taking the top link — is the term used by the Stanford History Education Group for a behaviour they observed distinguishing fact-checkers from the other groups. Stanford History Education Group, 'Evaluating Information: The Cornerstone of Civic Online Reasoning' (November 2016): 7,804 student responses collected across 12 states, finding widespread difficulty distinguishing sponsored content, evaluating evidence, and identifying who was behind online information. .org is an unrestricted top-level domain: registration is open to anyone, with no charitable, non-profit or organisational status required or checked. The padlock icon indicates an HTTPS connection — the traffic between browser and server is encrypted and authenticated to that server. It carries no information about the accuracy of the page's content, and is now present on the large majority of websites. The 'Institute for Everyday Nutrition' and its additive claim are invented for the worked example. No real organisation, claim or study is described there. The example is deliberately non-political: the episode's own argument is that a checking method used only on views you dislike stops being a checking method.