Check a Viral Claim in Five Minutes Before You Hit Share

The share button will still be there in five minutes. That's the entire trick, and almost nobody uses it.

What follows is a routine you can run on your phone, standing in line for coffee, on any claim that just made your blood pressure spike. It takes about as long as a song and a half. It works on a poll, a photo, a quote, a chart, a "BREAKING" post with 40,000 reposts and no link. And it works regardless of whether the claim flatters your side or attacks it, which is the only reason it's worth anything.

Minute one: find the thing itself

Most viral claims are a description of a document, not the document. Somebody read a study and tweeted about it. Somebody else screenshotted that tweet. A third account added red circles. By the time it reaches you, you're four steps from anything that can be checked.

So go upstream. Not to a different commentator. To the primary object.

If it's a bill, congress.gov has the text, and the summary at the top is usually two paragraphs. If it's a federal rule, it's in the Federal Register with a comment docket attached. If it's a Supreme Court opinion, the Court posts the PDF the morning it comes down, syllabus first. If it's a jobs number or an inflation number, bls.gov publishes the release on a schedule with the tables right there. If it's crime data, the FBI's Crime Data Explorer lets you pull it yourself.

You're not reading the whole thing. You're checking one sentence: does the document say what the post says it says?

Roughly a third of the time, it does. Another third, it says something close but weaker. The last third is the interesting one.

Minute two: check the date, then check the date again

Two dates matter and people confuse them constantly.

There's the date the event happened, and the date this particular post went up. A post from this morning can be describing a policy change from 2019, a lawsuit that was dismissed two years ago, or a riot in a country that isn't the one the caption names.

Old news going viral as new news isn't usually a lie. It's a re-run. Nobody technically fabricated anything, and that's exactly why it spreads so well. The caption says "this is happening right now," and the article underneath is real, and it's four years old, and the outcome everybody's furious about was resolved in court eighteen months ago.

Scroll to the byline date before you scroll to the comments.

Minute three: make the image prove itself

Long-press the image and run it through Google Lens, or upload it to TinEye. Both are free, both take about fifteen seconds, and both will show you where else that picture has lived on the internet.

The shark swimming down a flooded freeway comes back after every major storm. So does the same photo of a crowded shelter, the same burning car, the same wall of floodwater. Disaster photos are the most recycled content in the world because they're emotionally perfect and geographically anonymous.

Two other image tells worth knowing:

• The quote card. A politician's face, a plain background, a quotation, no source, no date, no outlet. If somebody actually said it on camera, a clip exists. If it was said in Congress, C-SPAN's video library is searchable and free. No clip, no transcript, no recording? Treat it as fiction until it isn't.
• The crop. A four-second clip of a nine-minute answer is a quote the speaker never gave. The fix is to find the full video and watch the ninety seconds on either side. Sometimes the context makes it worse. Fine. Now you know that, and you can say so with confidence.

Both parties' online operations run the cropped clip. It isn't one side's weapon. It's the standard equipment.

Minute four: pull the number apart

A poll result is four facts, and the headline only gives you one.

Look for who was polled, how many, when, and what exactly was asked. Reputable pollsters put all four in a PDF on their own site. Find it.

Who matters more than anything. A survey of all adults, of registered voters, and of likely voters will produce three different answers to the same question, and the gap isn't a rounding error. Somebody quoting an "all adults" number for a turnout question is telling you something true about the country and nothing useful about the election.

When matters because a poll taken before a debate is a historical document afterward.

What was asked is where the real mischief lives. "Do you support X?" and "Do you support X, even if it costs Y?" are different questions, and the second one isn't dishonest, it's just a different question. Read the actual wording. It's in the PDF, near the back, exactly as it was read aloud.

And margin of error. If a poll of 1,000 people has a margin around three points, the subgroup of 90 young men inside it has a margin you could drive a truck through. Any headline built on a crosstab of a hundred respondents is built on sand.

Minute five: the honesty check

Ask yourself one question, and answer it truthfully because nobody's listening.

Do I want this to be true?

If the answer is yes, raise your standard instead of lowering it. The claims that get smart people is the ones that confirm what they already believed, which is why smart people get got so often. Confirmation doesn't feel like bias. It feels like finally, someone said it.

The reverse test helps. Imagine the identical claim, same sourcing, same anonymous account, aimed at somebody you like. Would you repost it? If not, you don't believe the claim. You like the target.

When five minutes isn't enough

Sometimes you'll hit a wall. The study is paywalled, the original account is deleted, the video is grainy and nobody can identify the street. That's a real outcome, not a failure.

You've got three moves. Don't share it. Share it with the uncertainty stated out loud, in your own words, at the top. Or set it down and check again tomorrow, when the boring follow-up reporting has landed.

The Wayback Machine at web.archive.org handles the deleted-source problem more often than you'd expect. Paste the dead link, pick a date, read what was there.

If you already shared something wrong

Post the correction as a new post, not a quiet edit or a reply buried under forty comments. Same feed, same energy, same size. Say what you got wrong and what's actually true.

It costs you about thirty seconds of pride and buys you something that's getting genuinely scarce: people believe you the next time.

Your name is the only thing you're actually publishing. Everything else is somebody else's content, passing through.