ChatGPT Images 2.5 on Forgery Tasks: Testing Advertised Improvements Against Known Answers
Not on the map. Why?
Abstract (arXiv)
OpenAI released ChatGPT Images 2.5 on 8 September 2026, advertising more precise local edits, better consistency across edits, more faithful reference products and sharper detail. We evaluate these claims on four forgery tasks with answers fixed in advance: receipt-field alteration, repeated editing, product placement and small-print rendering. GPT-Image-2 provides same-week baselines at a cheaper and a more expensive tier. A limited improvement appears in receipt editing. After alignment, OCR detects changes to surrounding text in 31.7% of Flare outputs, against 44.2% for the cheaper baseline. This gain is concentrated on CORD receipts and sensitive to shifts of a pixel or less; the forged value itself is no more often correct. Repeated editing and fine print show no measurable gain. Product codes become more legible mainly because Images 2.5 draws the product larger. Defence outcomes change little: localisation remains weak for both generations. A detector that flags 68.6% of controlled benchmark images flags only 35.9% of images posted online. Advertised improvements therefore transfer unevenly to the tested forgery capabilities, while substantial detection limitations remain.
Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 18 of its 38 references that are in the map. First posted on arXiv 12 Sep 2026.
Builds on
18 of 38 references.
References · 18
- 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
- 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,764
- 2022DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generationcited by 3,286
- 2024Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preferencecited by 1,151
- 2024LLM Evaluators Recognize and Favor Their Own Generationscited by 651
- 2023Scaling Open-Vocabulary Object Detectioncited by 335
- 2025ImgEdit: A Unified Image Editing Dataset and Benchmarkcited by 225
- 2023TextDiffuser: Diffusion Models as Text Painterscited by 223
- 2023AnyText: Multilingual Visual Text Generation And Editingcited by 165
- 2022TruFor: Leveraging all-round clues for trustworthy image forgery detection and localizationcited by 130
- 2024AEROBLADE: Training-Free Detection of Latent Diffusion Images Using Autoencoder Reconstruction Errorcited by 123
- 2024Community Forensics: Using Thousands of Generators to Train Fake Image Detectorscited by 50
6 more
- 2021Spatial Dual-Modality Graph Reasoning for Key Information Extractioncited by 43
- 2024Revisiting Tampered Scene Text Detection in the Era of Generative AIcited by 21
- 2026GEditBench v2: A Human-Aligned Benchmark for General Image Editingcited by 11
- 2026GPT-Image-2 in the Wild: A Twitter Dataset of Self-Reported AI-Generated Images from the First Week of Deploymentcited by 8
- 2026AIForge-Doc: A Benchmark for Detecting AI-Forged Tampering in Financial and Form Documentscited by 5
- 2026When the Forger Is the Judge: GPT-Image-2 Cannot Recognize Its Own Faked Documentscited by 3
Built on by
Citing papers per year: 2026: 2.
2 papers.