Fake images are getting harder to spot. Here’s a handy guide.

Photographs have the power to shape our understanding of the world. And it’s never been more important to be able to distinguish between those that are authentic and those that are manipulated to advance an agenda, especially in the wake of dramatic or controversial moments.

But technological advances mean that identifying images that have been manipulated or even entirely generated by AI is becoming increasingly difficult. imaging becomes more and more delicate.

Consider a photo of Catherine, Princess of Wales, released by Kensington Palace in March. Media outlets removed it after experts noted obvious manipulation. Some have even questioned the authenticity of the images taken during the assassination attempt on former President Donald Trump.

Here are some expert tips the next time you come across an image that leaves you perplexed.

Enlarge

It may sound basic, but a study by researcher Sophie Nightingale of Lancaster University in Britain found that across all age groups, people who took the time to zoom in on photos and carefully examine different parts were better at spotting altered images.

Try it the next time you have a weird feeling about a photo. Just be careful not to focus on the wrong elements. To help you out, we created this (slightly exaggerated) example image to highlight some common signs of image manipulation.

Rather than focusing on things like shadows and lighting, Nightingale suggested looking at “photometric” clues like blurring around the edges of objects that might suggest they were added later; noticeable pixelation in some parts of an image but not others; and differences in coloration.

Consider this parrot: First of all, who brings a parrot into a polling station?

And take a closer look at its wings: the fuzzy edges of its main feathers contrast with the round cutouts closer to its body. This is clearly an amateur Photoshop job.

Looking for funky geometry

Fine details are some of the hardest things to edit smoothly in an image, which is why they often get missed. This is often easy to spot when regular, repeating patterns are disrupted or distorted.

In the image below, notice how the shapes of the bricks in the wall behind the divider are distorted and squashed. Something fishy has happened here.

Consider the now infamous photo of Princess Catherine.

The princess appeared with her two children in her arms. Online sleuths were quick to spot inconsistencies, including floor tiles that appear to overlap and a molding that appears to be misaligned.

In our polling station example, did you notice that this person had an extra finger? Of course, it’s possible that they have a condition like polydactyly, in which people are born with extra fingers or toes. This is a bit unlikely, though, so if you spot things like extra fingers, it could be a sign that AI was used to alter the image.

It’s not just bad Photoshop editing that ruins edits. AI is notoriously iffy when it comes to manipulating detailed images.

So far, this has been especially true for structures like the human hand, although progress is ongoing. Still, it’s not uncommon for AI-generated or edited images to show the wrong number of fingers.

Consider the context

One way to determine the authenticity of an image is to step back and consider what surrounds it. The context in which an image is placed can tell you a lot about the intention behind sharing it. Consider the social media post we created below for our edited image.

Ask yourself: Do you know the person who shared the photo? Are they associated with a post that seems intended to elicit an emotional response? What does the caption say, if anything?

Some doctored images, or even authentic images placed in a different context than reality, are meant to appeal to our “intuitive, visceral thinking,” says Peter Adams, senior vice president of research and design at the News Literacy Project, a communications research organization. non-profit organization that promotes critical evaluation of the media. These changes can artificially generate support or generate sympathy for specific causes.

Nightingale recommends asking yourself a few questions when you see an image that makes you react: “Why would someone post this? Is there an ulterior motive that might suggest it’s a fake?”

In many cases, Adams adds, comments or replies attached to the photo can reveal a fake for what it is.

Here’s a real-world example from X. An AI-generated image of Trump flanked by six young black men first appeared in October 2023, but reappeared in January, attached to a message indicating that the former president had stopped his motorcade to meet the men in an impromptu encounter.

But it didn’t take long for commentators to point out inconsistencies, such as the fact that Trump appeared to have only three large fingers on his right hand.

Go to the source

In some cases, authentic images appear out of nowhere and leave us wondering why they exist. Finding the source of these images can help us shed light on these facts.

Earlier this year, science educator Bill Nye appeared on the cover of Time Out New York wearing something a little more stylish than the blue lab coat many of us remember. Some wondered if the images were AI-generated, but tracing the credits back to the photographer’s Instagram account revealed that the scientist was wear young and bold clothes.

For images that claim to be from an actual news event, it’s also worth checking out news services like the Associated Press and Reuters and companies like Getty Images, all of which allow you to take a look at editorial images they’ve captured.

If you can find the original image, you have a genuine image.

Try a reverse image search

If an image doesn’t match the character depicting it, seems too biased, or doesn’t pass the mood test, image inversion tools like TinEye or Google Image Search can help you track down the originals. Even if they can’t, these tools can still bring valuable context to the image.

Here’s a recent example: Shortly after a 20-year-old gunman attempted to assassinate Trump, an image appeared on the Meta-owned social network Threads showing Secret Service agents smiling while clinging to the former president. The image was used to support the baseless theory that the shooting was staged.

People can use Google’s reverse image search to check the origin of an image and see if it has been manipulated or altered. (Video: The Washington Post)

The original photo does not contain a single visible smile.

Even armed with these tips, it’s unlikely that you’ll be able to tell real images from 100% manipulated ones. But that doesn’t mean you shouldn’t sharpen your skepticism. It’s part of the work we all need to do from time to time to remind ourselves that, even in times of division and confusion, factual truth still exists.

Losing sight of this, Nightingale said, only gives bad actors the opportunity to “throw it all away.”

“This is where society is really in danger,” she said.

Edited by Karly Domb Sadof and Yun-Hee Kim.