To check if a viral photo is real, find who posted it first, search for an older copy of the same image, look for reporting from people who were actually there, and run it through an AI image detector. That takes about five minutes. If the photo fails any of those checks, or you can’t finish them, don’t share it as fact.
Viral photos go wrong in two ways, and the fixes differ. Some are generated or edited with AI. Many more are real photos with a false caption: an old storm passed off as today’s, a protest from another country, a still from a film. An AI detector only helps with the first kind.
Why do fake viral photos spread so fast? #
Breaking news creates a gap. In the first hours after a disaster, protest or celebrity story, there are few verified images and huge demand for them. Engagement accounts fill that gap with whatever gets shares, and a dramatic generated image often outperforms a blurry real one.
Some of it is plain fraud. The FBI has warned that criminals create images of natural disasters and global conflicts to solicit donations to fake charities (FBI IC3).
A five-minute routine for any viral photo #
- Pause before sharing. Most of the damage happens in the first reshare.
- Find the first poster. Tap through to the account. Look for a named person or outlet with a history in that place. Aggregator accounts that repost everything are not a source.
- Reverse search the image. Google Lens, TinEye or Bing Visual Search will surface older copies. If the same photo appeared in 2019, today’s caption is false.
- Check for reporting from the ground. Search the event plus the location. Wire services and local outlets usually publish verified photos within hours. If a dramatic image appears nowhere in their coverage, ask why.
- Run an AI check. Save the image, or share it straight to a detector, and see what it says.
- Read the community context. Replies, quote posts and fact-check notes often flag fakes early. Treat them as leads to check, not verdicts.
How to check the photo itself for AI #
Save the highest-resolution version you can. Many apps let you open an image full screen and save it, which beats a screenshot because a screenshot lowers resolution and adds interface around the picture.
Then look at the image with the scene in mind. Viral fakes of disasters and crowds often fail on details a generator doesn’t know: signs in the wrong language, emergency vehicles with made-up markings, floodwater reflecting a different sky, faces that dissolve at the edges of a crowd. Our guide on how to tell if an image is AI-generated covers the full list of checks.
For a quick machine read, Expose AI sits in your phone’s share sheet. Long-press the photo in X, Instagram or your browser, share it to the app, and it returns Likely real, Uncertain or Likely AI-generated with a confidence meter, analyzed on your phone. If all you have is a screenshot, its Crop tool trims the app interface so only the picture is judged. It checks single images only, so a viral video needs other tools.
What if the photo is real but the caption is fake? #
This is the most common case, and no AI detector will catch it. A detector will correctly say “Likely real” about a genuine 2017 hurricane photo that someone captioned as last night’s storm. That’s why reverse search comes before the AI check in the routine above.
Clues that a real photo is being recycled:
- The weather, season or foliage doesn’t match the claimed date.
- Phones, cars or storefronts look several years old.
- The same image appears with different captions in different languages.
- The earliest uploader is on the other side of the world from the claimed location.
Signals that should stop you from sharing #
| Signal | Why it matters |
|---|---|
| No traceable original poster | You can’t know where or when it was taken |
| Older copies online | A real photo with a new, false story |
| Detector says Likely AI-generated | Strong sign of a generated image |
| Detector says Uncertain | Not enough evidence either way |
| Only engagement accounts carry it | Real events get covered by real outlets |
| The caption asks you to donate or click | A classic scam pattern |
“Uncertain” is not a pass. It means the image didn’t carry enough signal for a confident call, often because it has been compressed through several reposts. We explain what each result means in AI image checker results explained.
How to share responsibly when you can’t verify #
If you want to talk about an image you can’t verify, say so plainly: “Unverified, I can’t find the original.” Don’t crop out watermarks or platform labels. And if you shared something that turned out to be fake, a follow-up post correcting it reaches more of the same people than a quiet delete.
Frequently asked questions #
How can I tell if a news photo is AI-generated? #
Check whether a known outlet or wire service published it with a photographer’s name, then search for other photos of the same moment from different angles. Real events usually produce many images from many people. A single dramatic shot that nobody else captured deserves an AI check and extra caution.
Can a real photo be flagged as AI? #
Yes. Heavy filters, beauty modes, aggressive compression and some phone processing can push a detector toward “AI.” That’s one reason a detector result shouldn’t be the only thing you rely on. See why real photos get flagged as AI for the usual causes.
Do fact-checkers use AI detectors? #
Many do, alongside reverse search, geolocation and contacting the people involved. They treat a detector as one input because it can be wrong, and most fact-checks rest on finding the original source.
What should I do if I already shared a fake photo? #
Post a correction where you shared it, reply to anyone who reshared it from you, and then delete or edit the original post. Corrections work best when they say specifically what was wrong.