Atlas Ahead

How to tell whether you can trust what you just read

Every story on this site asks you to believe something. We try to earn it the same way each time: the claim goes in the sentence, the evidence goes in the link, and you can leave and check. That only works if checking is something you actually do.

Most of the time nobody does. Researchers analysed more than 35 million public Facebook posts carrying links, shared between 2017 and 2020, and found that around 75% of those links were shared without the sharer ever clicking through. Three quarters of the time, the headline was the whole of what got passed on.

That is not a story about other people being careless. Checking takes time nobody has budgeted, and most of the advice about how to do it is wrong—it tells you to study the page in front of you, which is the one thing the people who are good at this refuse to do. When Stanford researchers watched professional fact-checkers, history PhDs and undergraduates evaluate unfamiliar websites, the fact-checkers reached better-founded conclusions in a fraction of the time, and they did it by reading less.

This section is what they do, and where the limits of doing it yourself are. There is a real answer to the question of how much you should verify personally and how much you should hand to a newsroom that employs people to do it full time, and the answer is not “check everything”. Nobody checks everything. The skill is knowing which thirty seconds to spend.

That is the first thing you can do here, and it is the one this desk has the least right to ask you to skip: check us. What else belongs on this page—where to report what you find, how to back the people doing the original work—goes up when there is a real route to point at, and not before.

  1. Why a Source Is the Whole ClaimA sentence with nothing behind it is not information yet. It is somebody's assertion, formatted to look like more.
  2. The Ninety-Second CheckWhat professional fact-checkers do differently. It is faster than what you are doing now, not slower.
  3. When to Let Someone Else Do the CheckingYou cannot verify everything, and you were never supposed to. How to decide what to hand off, and to whom.
  4. How a Leak Becomes a StoryEleven and a half million files, a year of work, and why the delay was the point.
  5. What a Correction Tells YouThe outlets that correct in public are the ones keeping score. A clean record usually means nobody is.
  6. When the Number Is Doing the LyingTwo honest figures can disagree. Averaging them invents a third that nobody measured.
  7. A Photograph Is a ClaimPictures assert things sentences would have to defend. Most of the time the image is real and the claim is not.

Why a Corruption Desk Is Writing This

Our whole method is to point at somebody else’s reporting and say: this was established, here is who established it, here is what happened afterwards. A reader who takes that on trust is doing the thing this section argues against.

So read us the way the rest of this section tells you to read anyone. Every claim on this site links to the record it came from. If a link does not say what we say it says, that is the kind of mistake we would rather hear about than survive—the rules we bind ourselves to include correcting in public, and the address for it is on the same page.