How to spot a fake photo online
A Labour MP has ended up in hot water after sharing a doctored image of Rishi Sunak serving what looks like a terribly poured pint.
Although laughing at poorly poured pint is all fun and games the debacle has highlighted the dangers of doctored images being shared online and people believing them as real photos.
The problem is becoming even more dangerous with the growth in artificial intelligence generating convincing images of real people from scratch.
Karl Turner MP apologised after others pointed out the picture he shared on his Twitter account was fake with a community note later added to the post pointing out the truth.
When asked about it Turner said told Sky News he was “sorry for inadvertently sharing an image which apparently turns out to have been fake but the image looked a lot like what you would expect from useless out-of-touch billionaire Sunak.”
The original picture was shared by No10 and on Sunak’s Twitter account during a visit to the London beer festival to discuss the change in alcohol duty.
While there he was pictured pouring a pint of Black Dub from Wensleydale brewery which most punters would be happy with.
So how do you spot fake images online?
Check the focus, lighting and resolution
The first thing to do if you think an image is fake is to look for the telltale signs it’s been doctored.
In the Sunak picture above, if you look at it closely, it becomes quite clear the pint in his hand and the eyes on the woman behind him are not in the same resolution as the rest of the image.
They appear more blurry than the rest of the photo and the light on the doctored pint is hitting it at a different angle to the real image.
All of this combined makes it look like the pint in the doctored image does not quite fit with the rest of the photo.
If you ever have that thought about an object that doesn’t quite fit in an image it might be worth double checking it’s not fake.
Reverse image search
If you do have your doubts you can use reverse image search engines to find other instances of the same photo online.
If the search comes back with a near-identical photo except for a few subtle changes then one of the images may have been tampered with.
Some image search engines like Imagedited are specifically tailored to check to see if the photo has been doctored.
Check the metadata
Every digital image contains metadata, tiny bits of info that describe and categorise the photo.
It’s a sort of digital fingerprint that can tell you a lot, a legitimate photo will often have a proper file name, the date it was taken as well as information about the camera used to take it.
A fake image will often have next to no metadata and will probably have a file name that is irrelevant to the photo.
Metadata is editable so someone who is really looking to deceive an audience may make change it to be more convincing.
Social media companies often strip metadata from images for privacy reasons, so check to see if it has been uploaded elsewhere.
Was it generated by AI?
The growth in artificial intelligence (AI)-generated images can make some of these previous tips redundant, if a robot has made the photo then it is likely the lighting and resolution will be uniform.
However, at the moment AI technology isn’t perfect and although there are many different bots making images they all share a few flaws.
Firstly, AI technology struggles with hands. It doesn’t get them wrong every time but when trying to generate fingers in anything other than a relaxed position they often get confused and struggle to make an accurate human hand.
The same is true for jewellery and glasses, AI hasn’t quite figured out how they sit on a person just yet.
AI also tends to generate overly smooth surfaces often leading to pictures of people who appear to have perfect skin without any blemishes, dimples or wrinkles which we all have.
All of these tips can’t be applied to every AI-generated image and some of the best are almost impossible to distinguish from real life.
But several companies are building their own tools, including AI technology-based ones, which can determine if an image is real or generated.
Google has also said it is planning to improve its image search tool so it can identify AI-generated photos.