Digital Literacy: How to Check a Claim in Ninety Seconds

Checklists taught the wrong skill For years students were taught to evaluate websites by inspecting them: does it look professional, is there an author, is the domain .org, are there citations. Research by the Stanford History Education Group tested this against how professional fact-checkers work and found the checklist approach performs badly. Every one of those signals is cheap to fake.

Fact-checkers do almost the opposite. They spend very little time on the page itself and immediately open other tabs to find out what the wider web says about the source. The technique is called lateral reading, and it is faster as well as more accurate.

Lateral reading in practice Land on an unfamiliar site making a claim. Instead of reading down, open a new tab and search the organisation's name — adding a word like "funding," "founder" or "criticism" if the first results are all their own pages.

Within a minute you usually learn whether it is a research institute, an advocacy group, a company's content marketing, or a single person with a template. That context changes how you read everything on the page, and you got it without reading the page.

Wikipedia is a legitimate first stop here. Not as a citable source, but as a fast orientation with references you can follow — which is exactly what professional checkers use it for.

Trace the claim to its origin Most alarming numbers online are several hops from their source. An article cites a blog post citing a press release describing a study. Each hop drops caveats.

Click through until you reach the original. Frequently the study is smaller than reported, measured something narrower, or explicitly says the result is preliminary. If there is no origin — if the trail ends in another article citing "experts" — you have found a claim with no evidence behind it, regardless of how many places repeat it.

Images and video Old images resurface constantly with new captions. A reverse image search takes seconds and often finds the same photograph published years earlier in a different country.

For video, check the details rather than the drama: signage, language, weather, vehicle plates, architecture. Recycled footage usually fails on something mundane.

Recognising engineered outrage Content designed to spread optimises for emotional response, because sharing is driven by feeling rather than accuracy. That gives you a useful internal signal: the stronger your immediate urge to share something, the more warranted a ninety-second check.

Also watch for claims with no falsifiable content — vague authorities, unnamed studies, screenshots of deleted posts. Screenshots are trivially editable and carry no provenance.

Generated content Text and images from generative models are now common and often unlabelled. Detection tools are unreliable in both directions, so do not rely on them.

Provenance is the better defence. Ask who published this, whether they exist elsewhere, whether the claim appears in sources with editorial accountability. A well-written article from an outlet with no history, no named staff and no corrections policy deserves scepticism whether a human or a model wrote it.

The habit worth building You do not need to verify everything. You need a trigger. When a claim would change what you think, what you buy, or what you say to other people, spend ninety seconds: search the source laterally, find the original, check the date. That is a small tax on a small number of claims, and it filters out most of what would otherwise mislead you.