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

Yao Xiao, Binbin Yang, Weiyan Chen, Jiahao Chen, Zijie Cao, Ziyi Dong, Xiangyang Ji, Liang Lin, Wei Ke, Pengxu Wei
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:68491-68512, 2025.

Abstract

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 https://github.com/Coxy7/AIGI-Detection-Quality-Paradox.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-xiao25g, title = {Are High-Quality {AI}-Generated Images More Difficult for Models to Detect?}, author = {Xiao, Yao and Yang, Binbin and Chen, Weiyan and Chen, Jiahao and Cao, Zijie and Dong, Ziyi and Ji, Xiangyang and Lin, Liang and Ke, Wei and Wei, Pengxu}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {68491--68512}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/xiao25g/xiao25g.pdf}, url = {https://proceedings.mlr.press/v267/xiao25g.html}, abstract = {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 https://github.com/Coxy7/AIGI-Detection-Quality-Paradox.} }
Endnote
%0 Conference Paper %T Are High-Quality AI-Generated Images More Difficult for Models to Detect? %A Yao Xiao %A Binbin Yang %A Weiyan Chen %A Jiahao Chen %A Zijie Cao %A Ziyi Dong %A Xiangyang Ji %A Liang Lin %A Wei Ke %A Pengxu Wei %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-xiao25g %I PMLR %P 68491--68512 %U https://proceedings.mlr.press/v267/xiao25g.html %V 267 %X 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 https://github.com/Coxy7/AIGI-Detection-Quality-Paradox.
APA
Xiao, Y., Yang, B., Chen, W., Chen, J., Cao, Z., Dong, Z., Ji, X., Lin, L., Ke, W. & Wei, P.. (2025). Are High-Quality AI-Generated Images More Difficult for Models to Detect?. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:68491-68512 Available from https://proceedings.mlr.press/v267/xiao25g.html.

Related Material