Detecting AI-generated images has become increasingly difficult as AI models improve at a rapid pace. Traditional signs that used to give away AI-generated images, such as warped hands and jumbled text, are now rare. As a result, AI-generated images are fooling people more than ever, leading to the spread of misinformation. However, there are still ways to identify AI-generated images, although it requires more effort than before.
AI image detectors, such as AI or Not, Hive Moderation, SDXL Detector, and Illuminarty, use computer vision to analyze pixel patterns and determine the likelihood of an image being AI-generated. While these detectors are not foolproof, they provide a good starting point for the average person to assess the authenticity of an image. Anatoly Kvitnitsky, CEO of AI or Not, claims that their platform achieves a 98 percent accuracy rate on average.
In our tests, AI or Not correctly identified AI-generated images with an 80 percent success rate, while Hive Moderation achieved a 90 percent overall success rate. SDXL Detector on Hugging Face correctly identified 70 percent of AI-generated images, and Illuminarty classified 50 percent of them as having a very low probability. These results demonstrate that AI detectors are mostly effective but not infallible.
In addition to AI detectors, there are other methods for detecting AI-generated images. One approach is to use reverse image search tools, like Google Images, to trace the provenance of an image. This can help determine if an image is fake or if the depicted event did not happen. Google Search also offers an “About this Image” feature that provides contextual information about an image.
While AI-generated images are becoming highly realistic, there are still telltale signs that the naked eye can spot. These include warped objects, flawless skin, garbled text, and small inconsistencies like extra fingers or asymmetrical features. Looking for odd details, like stray pixels or subtly mismatched earrings, can also help identify AI-generated images.
Ultimately, detecting AI-generated images requires AI literacy and critical thinking skills. It is important to verify the source and context of an image, question the accompanying text, and consider the emotional impact of the image. Taking the time to evaluate the content before sharing or believing it is crucial in the age of AI-generated media.
To combat the spread of AI-generated deepfakes and misinformation, initiatives like the Coalition for Content Provenance and Authenticity (C2PA) have been established. C2PA offers clickable Content Credentials for identifying the provenance of images and whether they are AI-generated. On the other hand, the Starling Lab at Stanford University focuses on authenticating real images, particularly sensitive records like documentation of human rights violations.
While AI-generated images have various applications, they also pose challenges in terms of trust and transparency. It is up to individuals to be vigilant in detecting AI-generated images and to rely on a combination of methods, including AI detectors, reverse image search, and critical thinking, to assess the authenticity of images in the digital age.
