AI Image Detection: How to Detect AI-Generated Images

After the AI boom, the internet is flooded with AI-generated images, and there are very few ways for users to detect AI images. Platforms like Facebook, Instagram, and X (Twitter) have not yet started labeling AI-generated images and it may be a major concern for proving the veracity of digital art in the coming days. Thankfully, there are some easy ways to detect AI-generated images, so let’s look into them.

1. Detect AI Images Using Content Credentials

First up, C2PA has come up with a Content Credentials tool to inspect and detect AI-generated images.

Adobe, Microsoft, OpenAI, and other companies now support the C2PA (Coalition for Content Provenance and Authenticity) standard that is used for detecting AI-generated images. Based on C2PA specifications, the Content Credentials tool has been developed and allows you to upload images and check their authenticity.

Apart from images, you can also upload AI-generated videos, audio files, and PDF files to check how the content was generated. Here is how it works.

• To detect AI images, head to contentcredentials.org/verify.
• To check content, you don’t have to sign up or create an account.
• Next, click on “Select a file from your device” and upload an image.

• First, I generated an image using Adobe Firefly and uploaded it on Content Credentials. It correctly identified saying “This image was generated with an AI tool”.

• Next, I uploaded an image generated using OpenAI’s Dall-E 3 model on ChatGPT. And again, Content Credentials detected the image and said it was processed on ChatGPT.

• After that, I uploaded an image generated from Microsoft Designer (Bing Image Creator). It correctly identified the image as AI-generated.

Keep in mind that often you may get a “No Content Credential” or “Content Credential can’t be viewed” error if it’s a screenshot of an AI image or the image has been downloaded from social media, web, or even WhatsApp. These services remove the metadata or the image has been cropped, edited, or tampered with.

In that case, click on “Search for possible matches“. It will find the original AI image and you can verify all the changes then and there. I cropped and modified an AI-generated image and yet, it could find the original AI image.

Having said that, in my testing, images generated using Google Imagen, Meta, Midjourney, and also Stable Diffusion didn’t show any metadata on Content Credentials. More companies need to support the C2PA standard immediately to make it easier for users to spot AI-created pictures and stop the spread of digital deepfakes.

2. How to Check Watermark in AI Images

Some of the companies now add a watermark to AI-generated images. So check for watermarks around the corners of the AI-generated images. Of course, users can crop out the watermark, in that case, use the Content Credentials service and click on “Search for possible matches” to detect AI-generated images.

3, Check for Artifacts and Inconsistencies in AI Art

AI images generally have inconsistencies and anomalies, especially in images of humans. Pay close attention to hands, especially fingers as most current-gen AI models struggle to get that right. There are high chances of unusual distortion of hands, fingers, face, eyes, and hair. Also, look for the background and you might see complex patterns that may seem out of place.

Apart from that, zoom into images and look for digital artifacts. There could be strange pixelation, smudging effects, and high smoothening effects. You can also check shadow and lighting as AI image synthesis models often struggle to correctly render shadows and lights, matching the light source with the overall match.

So these are the few ways you can use AI image detection tools to verify the provenance of AI-generated images. While Google is working on its own SynthID for an invisible watermarking solution, it’s not available to people at large. Currently, it’s only available to limited Vertex AI customers.

Finally, OpenAI is also working with C2PA to develop and improve a robust standard for digital content certification. Anyway, that is all from us. If you have any questions, let us know in the comment section below.