Verifying what you see online

When anyone can generate convincing text, images and video in seconds, the question stops being can this be faked, because the answer is always yes. The useful question becomes how do I check. Professional fact checkers are not smarter than you. They just have habits, and the habits are learnable in an afternoon.

Read laterally, not deeper

The single biggest finding from research on fact checking is that experts leave the page. Instead of studying a suspicious article or video more closely, they open new tabs and check what the rest of the internet says about the source before engaging with the content. Who runs this site? What do others say about it? Has this claim been checked? Amateurs stare at the page and judge its look and tone, which is exactly what a faker controls. This habit is called lateral reading, and it beats any amount of squinting.

The research behind this is worth knowing. Researchers at Stanford compared professional fact checkers with historians and university students, and the fact checkers reached sound judgements about unfamiliar websites in a fraction of the time, precisely because they left the page early while the others stayed and studied it. And the reason the page itself cannot be trusted is simple once said: everything on the page is controlled by whoever made it. The clean design, the confident tone, the About page, the official looking logo, all of it costs almost nothing to produce, so none of it counts as evidence. What a faker cannot easily control is what the rest of the internet says about them, which is exactly where lateral reading looks.

Check the image, then check the story

For images, a reverse image search takes seconds: upload the image or paste its address into a search engine's image search and see where else it has appeared. A photo captioned as breaking news from this week that turns out to be years old from a different country is one of the most common forms of misinformation, no AI required.

Here is what the whole routine looks like on a typical case. A photo lands in the group chat showing flooded streets, captioned as a Queensland town this morning, and it is spreading fast. You search a few words from the caption in a news search and find no coverage, which is odd for dramatic flooding. You run a reverse image search, and the same photo appears in articles from years earlier about flooding in another country. Total time, about 90 seconds, and the verdict is solid: real photo, false caption. That combination, a genuine image moved to a new time and place, is so common because the image itself passes every visual check. It really is a flood. It just is not the flood the caption claims.

For big claims, apply the coverage test. If a story were true and important, reputable outlets with editors and legal teams would be reporting it. If a dramatic claim exists only on one account, one screenshot or one forwarded message, that absence is information. Real major news does not stay exclusive to a stranger's feed.

Why detection tools cannot do this for you

A misconception to clear before it costs you: surely there is a tool that can tell whether something is AI generated, so verification is really a software problem. The detection tools that exist are unreliable in both directions, missing polished fakes and flagging genuine content, and they lose ground as generators improve, for the same arms race reason covered in the deepfakes lesson. Provenance, meaning the traceable origin of a piece of media, is the check that holds up, because where something came from is a fact in the world that no generator can rewrite. Habits beat tools here, and the habits are already in this lesson.

Emotion is the tell

Content that makes you furious or triumphant spreads faster than careful content, and the people who design misinformation know it. A strong emotional spike is precisely the moment your checking instincts switch off, which makes the feeling itself a useful alarm. When something online makes your blood boil, treat that reaction as a signal to verify, not a signal to share.

There is solid evidence behind this. A large study that traced millions of posts on social media found false news spread further and faster than true news, and people, not automated accounts, did most of the spreading, apparently because false stories are engineered to be more novel and more emotionally charged than reality usually manages. So the outrage you feel is often the product working as designed, and sharing while you feel it is doing the designer's job for them.

None of this needs to take long. Lateral reading is a minute. A reverse image search is seconds. The coverage test is one search. The only genuinely hard part is the pause before sharing, because sharing feels urgent and checking does not. Build the pause and the rest follows.

Practise on the small stuff

In your own week these tests apply to smaller things than national news. The screenshot claiming your school is changing its uniform policy, the post that says a concert has been cancelled, the trade rumour about your team, the message that a paid app has gone free for 24 hours. Low stakes claims are the ideal practice ground, because the habit you build on trivial things is the one that will show up when a claim actually matters. Check one small thing a day and the checking stops feeling like work.

This lesson closes the loop on the whole topic. The deepfakes lesson showed that media can be fabricated, the scams lesson showed those fabrications aimed at your money, and the data lesson covered what you reveal that makes the targeting personal. Verification is the working answer to all three, and the case study that follows shows every one of these threads woven into a single phone call to a grandmother in Geelong. The habits do not make you a cynic who believes nothing. They make you a reader who knows the difference between what has been checked and what has not.