How to Tell AI Images Apart from Real Pictures

In this day and age, AI images are significantly more convincing and accurate than they were just a couple of years ago. That raises a question, how can we differentiate AI images from real pictures? Well, there are a couple of ways to spot if an image was made by AI or not. Here are some of them:

Lighting:

For example, the lighting on this lion is very dramatic.

Lighting in AI images are especially unrealistic and if you ask me it is the biggest give-away for AI images. AI tends to make the lighting of images more cinematic, the highlights yellower and the shadows bluer. In AI images the contrast is higher, the shadows are darker, and the color scheme of the image is more unified which does make it look “better” in our eyes but it’s also what makes it lose its realism.

Blurriness and Smoothness:

For example, she has perfect skin and her curls don't have a single stray hair, they're perfect! A bit too perfect, isn't it?

In AI images, parts of the image that are supposed to have texture or impurities often are unnaturally smooth and blurry. The image often feels “too perfect” because there’s no

imperfections like a stray hair or any blemishes on the skin. It feels fake and like that something is off. This is usually harder to spot since there are other ways to cover up any flaws in pictures but is still good proof that an image was generated by AI.

While some people do have 6 fingers, they show up way more in AI images!

Hands and Other Inconsistencies:

AI is notoriously bad at generating hands, usually messing up how many fingers there are. But there are some other things that AI gets wrong that are less obvious. For example;

A person could have three arms, their legs could turn into the legs of a table or a strand of hair could turn into an eyelash.

You can almost understand what's written, but it just doesn't make sense!

Text:

AI also struggles to generate text, making up or combining random words to make nonsensical ones that look close enough to a word if you aren’t paying attention. Letters are usually a little askew too, even if they form a word that’s logical.

Watermark:

An example of the watermark of an AI image, most AI watermarks are similar to this.

A lot of AI image generators put a watermark on the image it generates. For example, Gemini

puts a little star. If you spot any watermarks that aren’t for stock images or a watermark an artist has put, you can assume that the image was generated by AI. Unfortunately, these watermarks can be easily removed by either cropping the image or putting it into a watermark remover and some images don’t have them at all, but is undeniable proof of AI generation if found.

Without increasing the saturation and contrast, this image looks normal. Fascinating!

Hidden Digital Noise:

Because of the way AI generates images, if you saturate the colors of an AI image you’ll get glitchy and blocky patterns. To see this noise, you can go into any photo editing software and just increase the saturation and contrast. This is also indisputable evidence of AI generation, but takes some effort in order to prove so it isn’t the best way to tell images apart for daily use.

AI Detectors:

If you can’t be sure if an image was generated by AI or not, you can use an AI detector. AI detectors can be a useful tool to easily tell if an image was generated by AI but it can make mistakes too, thinking that some images that were made was made by AI. The images that the AI has been trained on are especially wrongly detected as AI too. For this reason, it’s best for you to look at an image first and then put it into as many AI detectors as you want.

Hopefully these tips were useful and that you learned something from reading this. I hope you have a great day!