AI detectors struggle to spot fake document images accurately
Beyond Natural Images: Rethinking AI-Generated Image Detection in Documents
Computer Vision and Pattern Recognition
Summary
Detecting images made by AI usually works well on photos, but not on document images like invoices or certificates. The authors found current detection tools perform worse on AI-generated documents because of unique features like uneven artifact patterns and text density. They created a new dataset of real and AI-made documents and showed that training detectors specifically on documents can help improve detection. This work is important for stopping fake documents from causing problems in sensitive areas.
What this means in practice
- •For document management teams: Improve tools that verify authenticity of scanned documents by integrating AI detection methods tailored for documents.
- •For financial fraud investigators: Enhance detection of forged financial paperwork by using document-specific AI image detectors that recognize subtle inconsistencies.
Authors
Zhangjie Fu, Jiazhen Yan, Yuanwen Chen, Xinquan Yu, Yanzhe Li, Hui Jiang, Lei Gao, Chenfu Bao
Abstract
AI-generated image detection has attracted increasing attention, but existing evaluations mainly focus on natural images, leaving AI-generated document images largely underexplored. This omission is concerning because documents often appear in sensitive real-world scenarios, such as invoices, expense reports, certificates, and medical records. In this paper, we first construct a controlled diagnostic benchmark, AIGDoc-Pilot, and reveal that existing detectors suffer substantial performance degradation on AI-generated document images, with the mean AUC dropping by more than 7%. Based on this, we further reveal two document-specific properties behind this gap: generation artifacts exhibit strong spatial inconsistency across local regions, and text density significantly affects real-synthetic separability, where text-dense regions offer stronger discriminative evidence. Motivated by these findings, we construct AIGDoc, a larger document-centric dataset containing diverse real-world documents and AI-generated counterparts produced by multiple advanced generation and editing models. Extensive experiments on AIGDoc demonstrate that existing detectors still struggle to reliably identify AI-generated documents, while document-based training partially narrows the gap. Together, these results offer valuable insights for developing dependable and generalizable detectors in document-centric scenarios. The code and datasets will be made publicly available upon acceptance of the paper.