Are High-Quality AI-Generated Images More Difficult for Models to Detect?

The remarkable evolution of generative models has enabled the generation of high-quality, visually attractive images, often perceptually indistinguishable from real photographs to human eyes. This has spurred significant attention on AI-generated image (AIGI) detection. Intuitively, higher image quality should increase detection difficulty. However, our systematic study on cutting-edge text-to-image generators reveals a counterintuitive finding: AIGIs with higher quality scores, as assessed by human preference models, tend to be more easily detected by existing models. To investigate this, we examine how the text prompts for generation and image characteristics influence both quality scores and detector accuracy. We observe that images from short prompts tend to achieve higher preference scores while being easier to detect. Furthermore, through clustering and regression analyses, we verify that image characteristics like saturation, contrast, and texture richness collectively impact both image quality and detector accuracy. Finally, we demonstrate that the performance of off-the-shelf detectors can be enhanced across diverse generators and datasets by selecting input patches based on the predicted scores of our regression models, thus substantiating the broader applicability of our findings. Code and data are available at \href{https://github.com/Coxy7/AIGI-Detection-Quality-Paradox}{GitHub}.

As AI-generated images become incredibly realistic, telling them apart from real photos is getting harder, posing challenges for content authenticity and security risks. This problem makes developing reliable AI image detectors crucial.Intriguingly, we've found a surprising twist: high-quality AI-generated images, which people usually prefer, are usually easier for current detectors to spot. To understand why, we explored what makes these high-quality images unique. Our research shows that images with certain visual traits, like high color saturation, strong contrast, rich textures, or simpler structures, tend to be both human-preferred and easier for AI detectors to identify.Our findings not only help explain this unexpected link between image quality and detectability but also show how we can use this knowledge. By focusing detection on parts of images that have these "easy-to-spot" characteristics, we can make existing AI image detectors perform better. This work guides future efforts to build more effective tools for identifying AI-generated content in the real world.