AI-Generated Image Detectors Overrely on Global Artifacts: Evidence from Inpainting Exchange

2026-01-30

AI Summary

Current AI-generated image detectors exhibit insufficient reliability in identifying localized inpainting, as they overly rely on global spectral artifacts introduced by VAE reconstruction rather than on genuine synthetic content. To address this, this work proposes the Inpainting Exchange (INP-X) operation, which preserves the inpainted region’s content while restoring original pixels in non-edited areas, thereby systematically revealing— for the first time—the detectors’ dependence on global artifacts. A newly constructed 90K test set based on INP-X demonstrates a significant performance drop in state-of-the-art detectors (e.g., from 91% to 55% accuracy), approaching random guessing. Conversely, models trained on this dataset show markedly improved generalization and localization capabilities, advancing the development of content-aware detection methodologies.

Technology Category

• Computer Vision: Generative Adversarial Networks (GANs) for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Search and Optimization: Local Search

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Modern deep learning-based inpainting enables realistic local image manipulation, raising critical challenges for reliable detection. However, we observe that current detectors primarily rely on global artifacts that appear as inpainting side effects, rather than on locally synthesized content. We show that this behavior occurs because VAE-based reconstruction induces a subtle but pervasive spectral shift across the entire image, including unedited regions. To isolate this effect, we introduce Inpainting Exchange (INP-X), an operation that restores original pixels outside the edited region while preserving all synthesized content. We create a 90K test dataset including real, inpainted, and exchanged images to evaluate this phenomenon. Under this intervention, pretrained state-of-the-art detectors, including commercial ones, exhibit a dramatic drop in accuracy (e.g., from 91\% to 55\%), frequently approaching chance level. We provide a theoretical analysis linking this behavior to high-frequency attenuation caused by VAE information bottlenecks. Our findings highlight the need for content-aware detection. Indeed, training on our dataset yields better generalization and localization than standard inpainting. Our dataset and code are publicly available at https://github.com/emirhanbilgic/INP-X.

Problem

• AI-generated image detection
• inpainting
• global artifacts
• content-aware detection
• image forensics

Innovation

• Inpainting Exchange
• AI-generated image detection
• global artifacts
• VAE spectral shift
• content-aware detection