Five minutes, five checks: how to check if a photo is real

You don’t need forensic software to check a suspicious photo. You need five minutes, three free tools and the discipline to stop when the answer is “unknown”.

Our guide on how to spot a fake image argued that the questions around a picture matter more than the pixels in it. This is the practical follow-up: what to actually do when an image lands in your feed and something about it bothers you. So if you want to know the easy way how to check if a photo is real, you have come to the right place.

This is part of a series of related articles om how to protect yourself from scamming on the internet.

The routine below takes about five minutes. It will not catch everything. It will catch the most common fakes, and it will stop you from becoming the next link in the chain.

Minute one: find the original

Most misleading images are real photographs with a false caption: an old flood presented as today’s, a protest from one country labelled as another. A reverse image search is the fastest way to catch them. It shows you where else the same picture has appeared, and often when.

Use at least two of these, because each one indexes a different part of the web:

• Google Lens. In a desktop browser, right-click the image and choose “Search with Google”. On a phone, take a screenshot, open the Google app and tap the Lens icon. Then open the About this image tab. It shows how other sites have used and described the picture, including news and fact-checking sites. It also shows any metadata the creator added and, for images made with Google’s own AI tools, whether it carries Google’s SynthID watermark (Google).
• TinEye. Upload the image or paste its address at tineye.com, then sort the results by Oldest. If a picture claiming to show yesterday’s event turns up years ago, you have your answer. One caveat: TinEye’s “first found” date is when its crawler found the image, not when it was first posted, so the real date may be earlier still (TinEye).
• Bing Visual Search. A useful third opinion, and sometimes better at finding near-copies that have been cropped or recoloured.

Two habits improve the results. Crop the image to its most distinctive part, such as a building or a face, because busy edges and added text confuse the matching. And look past the first page of results. The original is often a small local site, not the viral repost.

Minute two: read what the file says about itself

A digital image can carry information about where it came from: the camera, the date, the editing software, and increasingly whether AI was involved. Two cautions come first.

Metadata can be added or removed by whoever posts the image (Google). Most social platforms also strip much of it when a picture is uploaded. So missing metadata tells you almost nothing. Metadata that is present is a clue, not proof.

The stronger signal is Content Credentials. This is an open standard, known as C2PA, that attaches a cryptographically signed record to an image. It shows how the image was made and edited, and any tampering breaks the signature (C2PA). To check for one, upload the image to the free Verify tool at contentcredentials.org/verify. If a credential is there, you will see who signed it, which tools were used, and whether AI generation was declared.

Read the result honestly. A valid credential from a news agency or a camera maker is good evidence. No credential is the normal state of most images online, and says nothing either way.

Minute three: check the place

If the image claims to show a specific location, the location itself can confirm or sink the claim. This is the method open-source investigators use, and the basics need nothing more than a map.

• Signs and writing. Shop names, street signs, number plates and the language on posters narrow the country and often the city.
• Buildings and landmarks. Compare what you see with the same spot on Google Maps satellite view and Street View. A roofline, a bridge or a mountain on the horizon is hard to fake consistently.
• Light and weather. Shadows show roughly where the sun was. If the post says “this morning” and the shadows are long and falling east, something is off. Rain in the picture on a day the weather records show as dry is another warning.

This is often how fakes on X are caught. In one case described by researchers, a Community Note exposed an AI-generated image by comparing the building in it with satellite imagery of the place it claimed to show (Chrysidis et al., 2026).

Minute four: check the crowd

You are rarely the first person to doubt an image. Look for the people who got there before you.

• Community Notes. On X, a note can be attached to an image rather than a single post. It then appears automatically on other posts that use the same picture (TechCrunch). If a note is there, read its sources rather than just its verdict.
• Replies and quote posts. Look for people who say they were there, or who post the original. Treat both with the same care as the image itself.
• Fact-checkers. Search the claim in words, plus “fact check”. Established fact-checking organisations often cover viral images within hours.
• Independent witnesses. Real events produce many pictures from many angles. A dramatic scene with only one image, and no local reporting, is itself a finding.

Community Notes have one weakness worth knowing: they arrive late. In a large 2026 study, the first note on an AI-generated image took a median of more than 11 hours. By then, most of the sharing has happened (Chrysidis et al., 2026).

Minute five: decide

At the end of five minutes you will be in one of three places.

1. Confirmed. You found the original source, it matches the claim, and independent witnesses agree. Share it if you want to, ideally with the source.
2. False. The image is old, from somewhere else, altered or generated. Don’t share it. If someone you know shared it, a private message usually works better than a public correction.
3. Unknown. Nothing conclusive either way. This is the most common result, and the hardest to accept. Unknown means don’t share yet.

That third outcome is the real skill. Automated detectors will not rescue you from it. The same 2026 study found that the best tools identified AI images correctly only about 70 per cent of the time, and that their hit rate fell as the image generators improved. “I couldn’t tell” is an honest answer. Sharing it anyway is where the damage is done.

If you want to go further

For readers who check images regularly, the free InVID-WeVerify browser extension bundles most of the steps above into one place. It was built with the news agency AFP for journalists and fact-checkers. It runs reverse searches across several engines at once, reads metadata and applies forensic filters that can show signs of editing. Some advanced tools are reserved for registered journalists and researchers (WeVerify).

It is a powerful tool, and it does not change the principle. The software answers “what is this file?” Only you can answer “should I pass it on?”

A worked example: how to check if a photo is real, 5 steps easy

Take the fake image of an explosion at the Pentagon that briefly moved US stock markets in May 2023, and run it through the routine.

1. Find the original. A reverse search finds no earlier source. The picture appeared from nowhere, on accounts with no connection to the scene.
2. Read the file. Nothing useful: no credential, no camera data. As expected, this proves nothing either way.
3. Check the place. Compare with any photo of the Pentagon and the building does not match. Investigators at Bellingcat also noted a fence that seemed to melt into the barriers (NPR).
4. Check the crowd. No other angles and no witnesses for what would have been one of the most photographed events of the year. Within hours, the Pentagon’s protection agency and the local fire department said there had been no explosion.
5. Decide. False, on three independent grounds.

Steps three and four settled it, and neither needed anything more than attention.

The short version of how to check if a photo is real

Find the original. Read the file. Check the place. Check the crowd. Then decide, and remember that “unknown” is a decision too.

Five minutes is not much to spend before putting your name behind a picture.

think-smarter.net is independent. For organisations that want structured training on information resilience, see gadvisory.net.