I’m not gonna lie – AI-generated images are getting harder and harder to spot. They’ve become super-realistic, sometimes almost impossible to tell them apart from real photos. Add our short attention span to the equation, and the fact that we don’t spend too much time analyzing – and you get fake news and viral “photos” that aren’t photos at all.
But regardless of how advanced AI generators are, human reasoning still has an advantage (at least for now). And I’m on a mission to help you quickly figure out what you’re looking at. In this article, I’ll share some telltale signs that an image is AI, and I’ll share some tools that will help you confirm it.
In this article, I won’t discuss whether or not AI will replace photography (and why it won’t). This is just about the methods of telling the two apart and why it matters.
First things first, let us start with some signs that might indicate something’s not quite right.
1. Gut feeling
The first and the absolutely essential telltale sign is your gut feeling. Something just doesn’t seem right and you can’t quite put your finger on it? Well, this alone means that it’s worth digging further. If something seems off, it most probably is (not just in AI vs. photography, but in life in general, if you ask me).
For example, in this article, I told you about a Millie Bobby Brown image that showed up on my Facebook feed. It was super-realistic, no way could you tell it was a fake… Unless your inner voice spoke up. And mine did. It was a fake lingerie image, and I’ll tell you more about how I found out later.
2. Square format
This one is pretty simple and pretty common. The default format in all AI generators is square. It seems to me that most users of AI tools don’t change it – so the majority of images end up being in square format. Of course, this isn’t 100% true for all AI images, but it’s just another sign that increases the chances you’re looking at one.
This is especially true for a series of images, like those viral ones of a man kicking an alligator. In images like these, you can play the game we all loved when we were kids: find the difference. :) Look at them, compare them, and think: are all the elements consistent? It requires some time, but it will tell you a lot!
Major AI generators like Midojurney have been notorious for creating weird human hands. Although this has improved a lot – it’s not perfect, at least not yet. So, take a closer look. Irregularities in hands can still reveal whether an image is AI or not.
5. Tech items/electronics
Tech items and electronics are among those things AI generators still can’t get right. Things like cameras, audio gear, keyboards, computers… All of them look completely unrealistic! It’s our nature to pay more attention to humans, especially their faces, in images. But if there’s any kind of electronics in the image, pay attention to that too. Chances are it will tell you the truth about the image you’re looking at.
I noticed another thing, at least with Midjourney – it sucks at creating cars. I mean, they look relatively realistic – but they aren’t real. All those cars you see in AI-generated images are car models that don’t exist.
Keep in mind, though, that this may not be something anyone can spot. You need to know cars really well to notice it, and it was a car enthusiast who drew my attention to it. But if you are among them, this can be another indicator of fake imagery.
7. Items connection
Here’s another sign that can be pretty obvious, but you need to take a closer look. Playing with AI generators, I noticed that all of them are pretty bad at generating a proper connection between two items, or two ends of the same item. For example, things like necklaces and camera straps nearly always look weird, and they’re not connected properly. The same goes for cables and wires.
Finally, AI generators are still pretty bad with text. If you spot weird letters, non-existent words, or sentences, there’s no doubt it’s AI. However, this is getting better, and it’s happening fast. So, this telltale sign may soon not be easy to spot… If possible at all.
How to check if it’s AI or a real image?
Okay, so you may have spotted some obvious telltale signs, and you’re sure the image you’re looking at is AI-generated. But what if all you have is your gut feeling and the image is flawlessly realistic? Well, there’s still a chance to check its authenticity. Below, you’ll find some that I use.
Google Image search
An image of Donald Trump’s arrest has popped up in your Facebook feed seven times, and you haven’t even had your morning coffee yet? Or maybe it was the pope wearing a puffy coat? With viral images like these, the truth is one Google Image Search away. I’ll let the examples speak for themselves:
Google recently introduced several new tools for checking whether an image is genuine or AI, how old it is, and where it had previously been published. This can also help you determine whether or not it was AI… Or if it was real, but used in a wrong context.
AI or Not website
There are quite a few websites that can tell you with high or relatively high accuracy whether an image was AI-generated or not. Probably the most popular is AI or not, and we wrote about it here. However, keep in mind that none of these tools is 100% accurate, so always keep your critical thinking activated. :)
Find the author
Finally, if you see an image online and your guts tell you something’s fishy about it – check for a watermark. Sometimes, a creator will be signed, and you’ll find a watermark with their name or nickname embedded in the image. This is what I did with a Millie Bobby Brown image, I wrote more about it here.
In short, I recently saw a very “spicy” image of the young actress. It was super-realistic, but it just wasn’t something I could believe was real. There was a watermark in the image, so I found the creator and noticed that all of their artwork was tagged with “celebrityfake” and “celebfake.” This takes a little more work, but it’s one of the techniques that can sometimes reveal the truth even behind super-believable AI images.
Why is this important?
I’ll put this briefly: you don’t want to be lied to and manipulated, do you? If the answer is no, this is why you need to distinguish between AI and real photos. I know, it’s getting more difficult. And I know, you don’t want to think too much while mindlessly scrolling Instagram and Facebook. I mean, that’s why it’s called “mindlessly scrolling,” right?
However, if an image causes a strong emotional reaction – then it’s time to turn your brain back on before you comment or share it. Be it laughter, disbelief, anger, or sadness; anything that fires up your emotions is worth questioning and double-checking in this mad world full of fake news. This way, you’ll avoid being fooled, manipulated, and lied to, both for silly things like the pope’s puffy coat, but also for way more serious topics like wars or terrorist attacks.
As always, I urge you to listen to your gut feeling, above all. AI images are getting more realistic, but your intuition will be there even when your brain is in the “off” mode while you’re killing time on Instagram. And if it speaks to you, I hope that the tools and telltale signs I gave you above will be helpful.