On this page
Why forwarded photos are hard to judge
A photo that reaches you through WhatsApp, Telegram or a family group has usually travelled a long way. Each hop compresses it, renames it and strips the details that once said where and when it was taken. By the time it lands in your chat, it often arrives with nothing but a caption, written by someone who wasn't there.
- The source is gone. You can see who sent it to you, not who took it.
- The date is gone. Messaging apps strip most of a file's details, so a picture from years ago looks as new as one taken this morning.
- The labels are gone. Content Credentials and the other labels AI tools write into a file rarely survive a forward, so "no AI label" tells you nothing.
- The caption does the work. A real photo with a false caption misleads just as well as an AI image.
On WhatsApp, a "Forwarded" label means the person who sent the message didn't write or take it, and a double-arrow "Forwarded many times" label means it has passed through a long chain of chats. Neither says the picture is false. Both mean your contact is not the source, and that nobody along the way had to check it.
The two-minute routine
You don't need special skills, just a few habits. Run through these steps before you forward anything that would upset, frighten or anger people if it were true. Most pictures take a couple of minutes; a few take longer, and those are the ones that most needed checking.
Pause
Strong emotion is the hook. Pictures made to deceive are built to make you feel first and think later: outrage, fear, pity, pride. If a picture makes you want to share it right away, that is the moment to slow down.
Find the source
Run a reverse image search with Google Lens or TinEye, and look for the first time the picture appeared and who posted it. WhatsApp has tested a "search on web" option for images; if your app shows it, it does the same job. Our reverse image search guide walks through each step.
Check the date and place
If the earliest copy is older than the event it claims to show, or comes from another country, you have your answer. Look in the picture for clues that fit the caption, or don't: the season, the language on signs, the cars, the uniforms, the weather.
Look for the same moment from others
Real events are photographed by many people. Search for the event in trusted news outlets. If a dramatic scene exists in exactly one picture, shared only in forwards and by accounts you can't identify, be careful.
Check the picture itself
Drop the picture into Perspicax. It examines every region of the image and answers likely AI-generated, likely real or inconclusive, and it works on the compressed copies that forwards leave behind.
Decide
If you can confirm the picture, share it with its source. If you can't, don't forward it. A picture you didn't pass on can't mislead anyone.
Signs a forward needs a closer look
Some things about a forward should make you reach for the routine straight away. Others feel reassuring but prove nothing at all.
Still worth checking
- No source, no date and no place: just a caption and a request to share.
- Urgency: "share before they delete it", "the news won't show you this".
- A caption that doesn't fit the picture: summer clothes in a winter story, signs in the wrong language, the wrong kind of police car.
- A dramatic scene that no news outlet or other witness seems to have photographed.
- A screenshot of a "news" graphic with no link to the outlet that supposedly published it.
No longer reliable
- Lots of people have shared it: forwards spread because they are striking, not because they are true.
- Someone you trust sent it: they most likely received it from someone else.
- It looks too real to be fake: the newest AI images look exactly like photos.
- There is no AI label: forwarding strips labels, so their absence proves nothing.
- There are no obvious mistakes: hands, faces and text are usually right in modern AI images.
None of these signs settles anything on its own. They tell you when the two-minute routine is worth running, which, for anything dramatic, is nearly always.
How Perspicax helps
Perspicax combines proven computer vision techniques with TrueEye 2.3, our state-of-the-art detector. It reads the picture itself, region by region, down to the pixels, so it doesn't need the labels and details that forwarding strips away. It answers in one of three bands:
Three possible answers
Forwarded images are hard for any detector: they are squeezed, resized and re-saved at every hop. That is why we test on images that have travelled. On RealChain, a public test set of images passed along real sharing chains, Perspicax scored 94.9% accuracy. On real photos shared on the web and social media, it wrongly flagged about 1 in 100, which is why an answer is evidence, not a final word.
Measured the way Perspicax answers: an inconclusive answer counts as a miss. See every number
To check a picture from your phone, save it from the chat or copy it, open perspicax.io in your browser, and paste or choose it. When you upload a screenshot, Perspicax crops to the photo inside it. Our guide to checking a photo on your phone covers the details.
What to do with the answer
The answer is one piece of evidence, read next to everything else you found:
- Likely AI-generated: don't forward it as real. If it is already spreading in your group, say what you found.
- Inconclusive: there was not enough evidence either way. Go back to the source and the date, and keep the picture to yourself until you know more.
- Likely real: the picture itself looks like a camera photo. That does not make the caption true, so check the date and the place before you share it.
A real photo can still be the wrong photo. Perspicax tells you about the picture, not about the story wrapped around it, which is why finding the source comes first in the routine.
How to correct a forward kindly
Sooner or later someone you love will share something false. A quick, kind correction works better than a public argument, and keeps the group willing to listen next time:
- Reply to the message itself, so the correction sits right next to the picture.
- Share what you found, not just a verdict: "this photo is from a flood in another country years ago, here is the original".
- Assume good faith. Most people forward things because they care, not because they want to mislead.
- In a large group, a private message to the sender can work better than a reply to everyone.
- Don't forward the false picture again, even to debunk it: describe it, or share the original instead.
If you shared something false yourself, say so in the same chat and delete the message if your app lets you. A quick correction travels some of the way the picture did.
When to take extra care
Some forwards do more harm than others. Slow down most with pictures that:
- show disasters, attacks, wars or accidents, where false pictures spread fastest and hurt the most;
- ask for money or donations, especially for a named person or a cause you can't check;
- show a public figure doing something shocking, often just before an election or a big decision;
- promote a cure, a product or an offer that seems too good to be true;
- warn about a local danger, such as a crime, a scam or a missing person, with no link to the police or a local news outlet.
For pictures of breaking news, our guide on how to verify a news photo goes further. For the pictures in your social feeds, see how to spot AI images on social media.
Check an image now
Drop, paste or link a photo. Perspicax checks every region of it with computer vision and TrueEye 2.3, and answers likely AI-generated, likely real or inconclusive.
Questions people ask
How can I tell if a WhatsApp photo is real?
Find where it came from with a reverse image search, check that the date and place fit the caption, look for the same moment in trusted news, and check the picture with Perspicax. If you can't confirm it, don't forward it.
What does "Forwarded many times" mean on WhatsApp?
That the message reached you through a long chain of chats, so the person who sent it to you didn't create it. It doesn't mean the picture is false, but it does mean its source is unknown.
Does a forwarded image keep its AI label?
Rarely. Messaging apps compress pictures and strip most of a file's details, including Content Credentials. A missing label proves nothing, so check the picture itself.
Can Perspicax check a compressed or screenshotted photo?
Yes. It examines the picture itself, region by region, so it works on re-shared copies, and a screenshot is cropped to the photo inside it. On RealChain, a public test of images passed along sharing chains, it scored 94.9% accuracy. The original file is still best when you can get it.
Is a photo real if a reverse image search finds nothing?
No. A brand-new picture, real or AI, has no history yet. No results means you need other checks, not that the picture is genuine or fake.
What should I do if I already shared something false?
Post a short correction in the same chat with what you found, and delete the message if your app lets you. The sooner you correct it, the less far it travels.