You will be able to trace an image or clip back to where it first appeared.
A photo arrives in Hafiz's condo group: a road under brown floodwater, cars half submerged, with the caption "Bedok North now. Avoid the area." Within ten minutes it's been forwarded to three other groups. Someone asks whether it's AI. Someone else says it can't be, it looks too real.
Both of them are asking the wrong question. Whether an image is AI-generated matters less than where it came from. This lesson shows you how to trace an image or clip back to its first appearance, which answers most of the questions that matter.
A reverse image search lets you search the web using an image instead of words. It shows where the same or similar images have appeared before, often with dates.
Two common tools are Google Lens, built into the Google app and Chrome, and TinEye, a website where you upload an image or paste its address. On a phone, you can usually save the image from WhatsApp, then open it in Google Lens or upload it to TinEye from your photo gallery. Try more than one tool, because each searches a different collection of images.
Hafiz saves the flood photo and runs it through Google Lens. The results show the same photo in a news article from several years earlier, about flooding in a different country. The cars, the shop signs in the background, everything matches. The caption is new. The photo isn't.
That outcome is far more common than a fully AI-generated image. When something dramatic happens, old photos and videos from other places and other years get recaptioned and shared as if they were happening now. It's quick, it needs no special tools, and the image is real, which makes it hard to dismiss by looking for flaws.
So when you check an image, the question to answer first is: when and where did this first appear? If the earliest appearance you can find is years before the event it supposedly shows, or in a different country, you have your answer, however real it looks.
Video works the same way. Take a screenshot of a clear frame, ideally one with distinctive details such as a sign, a building or a landmark, and search with that. Clips often turn out to be from old news reports, films or entirely different events.
Some AI tools and platforms now add information to images to show how they were made. Some attach content credentials: a record inside the file describing which tool created or edited it. Some platforms show a label such as "AI-generated" or "Made with AI" when they detect this information or when the uploader declares it. Some image generators also add invisible watermarks.
These are helpful when they're present. If an image carries a credential saying it came from an image generator, you've learned something real. But a missing label proves nothing. Credentials are often stripped when an image is screenshotted, compressed, or forwarded through messaging apps, and many tools don't add them at all. An image with no label could be AI-generated, a real photo, or a real photo with a fake caption.
So use labels as one clue among several, never as the deciding check.
You may have heard that AI images give themselves away with mistakes: hands with six fingers, garbled text on signs, earrings that don't match, odd reflections. That was often true of earlier tools, and you'll still see some of these errors.
But newer tools make far fewer of them, and the improvement keeps coming. Relying on glitches gives you two problems. You'll miss AI images that happen to have none, and you'll wrongly dismiss real photos that have an odd-looking detail, as real photos often do. A blurry hand in a genuine flood photo proves nothing either way.
Treat glitches the way you treat labels: if you spot an obvious one, it's a clue worth noting. If you don't, you haven't learned anything. The reliable check is still tracing the image to its source.
Sometimes a reverse image search turns up nothing. The image might be new, it might come from a closed group, or it might be AI-generated and posted for the first time. In that case, go back to lesson 6.2, Lateral reading: leave the page to check it. Search for the event itself. If there really were serious floods in Bedok North, the national water agency PUB, news outlets and other residents would be reporting it, often with their own photos.
If you still can't confirm it, the image is unconfirmed, and it shouldn't be shared as if it were true. Hafiz posted a reply in the condo group with a link to the original news article and a short note: "This photo is from an older flood overseas. Checked on Google Lens." A few people thanked him. Nobody forwarded it again.
You don't need to wait for the next flood photo to practise. Your chat groups almost certainly hold a few images from the past week that were passed around as news, and they're good material to start with.
Run a reverse image search on two images from your chat groups and record the earliest appearance you can find.
Junxiong-WFG Organisation is an authorised representative of AIA Financial Advisers Private Limited (Reg. No. 201715016G).