How to Spot Fake News: A Practical Guide

Not one thing: three kinds of toxic information

“Fake news” has become the catch-all label for anything unconvincing online, but UNESCO’s handbook for training journalists proposes a more precise term: information disorder, which covers far more than outright falsehoods. Much of the problematic content, explains researcher Claire Wardle (First Draft), isn’t even invented: it’s genuine material used out of context, and that kernel of truth is exactly what makes it more believable and more shareable.

Within that disorder, three categories stand apart:

Practical example: someone receives a screenshot of an article with a shocking headline in a chat. Before forwarding it, they open another tab and search for the same headline elsewhere: if no known outlet is reporting it, and the originating site has no verifiable newsroom behind it, that’s reason enough to stop.

How it spreads: social media, websites and word of mouth

The Reuters Institute’s 2026 survey shows how much this weighs on trust in the news: 62% of people worldwide say they’re concerned about fake news online, four percentage points higher than the year before, with increases above five points in eleven markets (the only exception is Brazil, down three points). 58% worry they can no longer tell what’s real from what’s false when reading news online.

On specific channels, the same survey records low trust: only 22% trust news they read on social media and 20% trust answers from an AI chatbot, against an overall average trust level of 37%, the lowest recorded since 2015. Yet in 2026, social media and video platforms overtook both television and news outlets’ own sites and apps as a news source for the first time, on average across the monitored markets: the most-used channel isn’t the one people trust most. AI chatbot use for news remains a minority habit but is growing, from 7% to 10% of users weekly, peaking at 16% among people under 35.

Alongside social platforms there are websites: pages that copy the logo and layout of real outlets (what Wardle calls “imposter content”), or stories invented from scratch and published on purpose-built domains (“fabricated content”). Understanding how the internet works and how you land on a site helps spot the details that give away a clone: an address that’s almost, but not quite, identical to the real one, with no verifiable information about who runs it.

Then there’s digital word of mouth. Many false stories don’t start with an organized plan — they start with ordinary people sharing an unverified item in good faith, in a family chat or a work group. As it passes from hand to hand, that story can change shape and turn into outright misinformation, even without anyone along the chain ever intending to deceive.

Why it goes viral (among the factors that matter)

A 2018 MIT study published in Science analyzed roughly 126,000 true and false stories shared on Twitter between 2006 and 2017, adding up to about 4.5 million shares by 3 million people. The result: falsehoods traveled farther, faster and more deeply than the truth, across every category of information tested, with an even stronger effect for political stories compared with terrorism, natural disasters, science, urban legends or financial information.

The numbers show the size of the gap: the top 1% of false-news cascades reached between 1,000 and 100,000 people, while true stories rarely crossed 1,000. True stories took six times longer to reach 1,500 people, and twenty times longer to be shared ten times from the original post; overall, a false story was 70% more likely to be retweeted than a true one.

Among the factors the authors identified is perceived novelty: false stories came across as more surprising than true ones, which appears to push people toward sharing them precisely because they’re unfamiliar. Emotional response matters too: falsehoods triggered more fear, disgust and surprise, while true stories triggered more anticipation. One finding rules out a convenient explanation, though: bots amplified true and false stories at roughly the same rate, so what mainly explains the faster spread of falsehoods is human behavior, not automation.

An even more effective method for spotting it

An IFLA infographic (based on a 2016 FactCheck.org article) lays out eight practical steps, designed for use in libraries: check the source by leaving the page to investigate it, verify the author is a real person, check the publication date, recognize your own biases, read past the headline, check whether the cited links actually support the story, work out whether it’s satire, and, when in doubt, ask a librarian or consult a fact-checking site.

Research on professional fact-checkers, conducted by the Stanford History Education Group and covered by Poynter, found an even more effective technique: lateral reading. Instead of staying on the suspicious page and judging it by its appearance — URL, design, functionality — readers open other browser tabs and check what independent sources say about the same site or story. Those “internal” cues, which feel like the most natural things to evaluate, turned out to be unreliable; what matters more is the trail a source leaves elsewhere, in how others have already judged it.

Photos and videos made with AI: what to look for

As generative AI has spread, Wardle’s “manipulated content” and “fabricated content” categories gained a new tool: images and videos built from scratch, increasingly hard to tell apart from a real photo. Facta.news, an Italian debunking project, suggests a few practical checks: look for watermarks from major generators (an animated cloud for Sora 2, a four-pointed star for Gemini, a logo for Grok); notice anomalies, such as poorly reproduced brand logos, garbled text, crowds that look too uniform, objects that appear or vanish, or impossible perspectives; and be wary of suspicious “perfection,” such as skin that’s too smooth, teeth that are too even, no background noise, or motion that’s too fluid.

One extra trick: people who create disinformation often deliberately lower the resolution of their content, making it harder to spot the details that would give away a composite image or a synthetic video. That’s why the strongest check remains the same one that works for text: trace the content back to its original source (creators of AI-generated material often disclose it openly on their own profile) and compare the event it depicts against what reliable news outlets are reporting. What decides what we see online in the first place isn’t neutral either — the same goes for the algorithms that decide what shows up first on a social feed, one more reason not to stop at the first thing that appears in your timeline.