How do you check a viral photo or video before sharing it? Find where it first appeared, check whether the same picture exists from an earlier date or a different place, and look for independent confirmation from people who were actually there. Spotting AI-generated details helps, but most misleading content is real footage with a false caption, so source and context matter more than pixels.
The routine below takes five to ten minutes with free tools on a phone. It will not catch everything, but it catches the most common tricks: old images recycled for new events, footage from another country, misleading crops and AI images presented as photographs.
Why are most misleading posts not deepfakes?
When people think about fake media they picture sophisticated AI. In practice, a large share of misleading posts during floods, storms, accidents or protests are genuine photos and videos from another time or place, given a new caption. They are cheap to make and hard to dismiss, because everything in the picture is real.
The second common type is the selective edit: a crop that removes context, a clip that starts after the provocation, or two videos stitched together to suggest cause and effect. The third type is the fully AI-generated image, which is now easy to produce and is often posted first as a joke or as art, then reshared by others without that label.
This matters for how you check. Looking for AI artefacts such as odd fingers or garbled text can catch some synthetic images, but it says nothing about a real photo with the wrong caption. The questions that cover all three types are the same: who first posted this, when, where, and does anyone independent confirm it?
Why speed and closed groups make checking harder
Speed is the enemy. Misleading posts spread fastest in the first hours of a breaking event, exactly when official information is thin. Waiting an hour before sharing costs almost nothing, and many viral claims are debunked within that hour.
There is also a practical Vietnamese angle: a lot of viral content travels through closed Zalo and Facebook groups, where screenshots of screenshots have lost all metadata and source links. In those cases, simply asking “where did this come from?” in the group itself is the most valuable check you can make.
Step-by-step: how to verify a viral image
Work through these steps in order and stop as soon as you reach a clear answer.
- Pause and restate the claim. Write down in one sentence what the post says happened, where and when. The claim is often vaguer than it felt when you first saw it.
- Run a reverse image search. Use Google Lens, Bing Visual Search or a similar tool on the image or a screenshot of it. Look for the earliest dated copy and for versions in other languages or countries.
- Search with different crops. If nothing turns up, crop to a distinctive part of the picture, such as a building, the shape of a sign or a vehicle, and search again. Cropped images often defeat simple matching.
- Find the original uploader. Follow the shares back to the first account that posted it. An eyewitness with other photos from the same place and time is a good sign; a page that reposts viral content all day is not.
- Match the location. Compare landmarks, road layout, vegetation and weather with map imagery or other photos of the claimed place. A “storm in central Vietnam” photo with bare winter trees is a clue.
- Look for independent confirmation. Check whether reputable news outlets, local authorities or several unconnected people describe the same event. One source repeated a hundred times is still one source.
- Decide and label honestly. If you cannot confirm it, do not share it, or share it with a clear note that it is unverified. If you shared something false, correct it publicly.
How to check a video and spot AI-generated images
Grab keyframes. Pause the video at clear moments, take screenshots and reverse-search them like photos. The free InVID-WeVerify browser plugin, widely used by fact-checkers, can extract keyframes for you.
Listen to the audio. Language, accent, sirens and public announcements can reveal the country. Audio replaced during editing often has a different ambience from the scene on screen.
Check the weather and light. Shadows, rain and daylight should fit the claimed time. A video said to be from this morning with long evening shadows deserves a closer look.
Look at hands, lettering and repeated patterns. AI images still sometimes produce odd fingers, melted lettering on signs, repeating tiles or distorted faces in the background. These are hints, and they are disappearing fast.
Be wary of images that are too perfect. Waxy smooth skin, cinematic lighting at a disaster scene and perfectly composed “news” shots can point to generation. Real breaking-news photos are usually messy.
Check for content credentials. Some cameras, editing apps and AI tools now attach provenance information following the C2PA standard. When present it can show how an image was made, but its absence proves nothing, because social platforms often strip it.
Read metadata carefully. Original files may carry EXIF data with the camera model and date, but screenshots and most social uploads remove it, and it can be edited. Treat metadata as a supporting clue only.
Use AI detectors with caution. Online “is this AI?” detectors give probabilities and are often wrong on compressed or edited images. Never treat a single detector score as proof.
When should you stop checking and simply not share?
Some posts cannot be verified quickly, and that is an acceptable outcome. If an image involves an accident, a crime or a child, the bar should be higher: sharing an unverified picture can expose real people to harassment even if it turns out to be accurate.
Be wary of posts that come with a call to action: share before it is deleted, donate to this account, avoid this area. Urgency combined with a request is a pattern worth slowing down for, and donations should go through organisations you can verify independently.
In family groups, a gentle reply works better than an accusation. Something like “I searched this picture and found it came from a different event”, with a link, lets people save face and still learn the method.
If you run a business page or a community group, give admins one simple rule: no breaking-news images without a source link. It protects the page’s credibility, which is slow to build and quick to lose.
Frequently asked questions
What is the fastest way to check if a photo is old?
Run a reverse image search with Google Lens or Bing and look for the earliest dated copy. If the same image appeared years earlier, the new caption is false.
Can AI-generated images always be detected?
No. Visual flaws are disappearing and detectors make mistakes, especially on compressed images. Checking the source and context is more reliable than judging the pixels.
Does missing metadata mean an image is fake?
No. Most social networks and all screenshots remove metadata automatically. Missing metadata is normal and proves nothing either way.
What are content credentials?
They are provenance records, based on the C2PA standard, that some cameras, apps and AI tools attach to files to show how they were created and edited. They help when present, but many images online do not carry them.
How do I tell relatives a post is false without upsetting them?
Share what you found rather than a verdict, for example the earlier version of the photo with a link. Showing the method lets them check for themselves next time.
Should I share a post with a note saying it is unverified?
Only if sharing serves a clear purpose, such as asking people with local knowledge to confirm. For alarming or personal content, not sharing is usually the better choice.
