# Can You Trust What You See? Human and AI Detection of Synthetic Legal Evidence

arXiv: https://arxiv.org/abs/2606.07613 · HTML: https://mlsanity.com/paper/2606.07613/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 21 of its 52 references that are in the map.
- First posted on arXiv 29 May 2026.

## Builds on (most cited first, 21 of 21)

- [Hierarchical Text-Conditional Image Generation with CLIP Latents](https://mlsanity.com/paper/2204.06125/) (2022)
- [A Style-Based Generator Architecture for Generative Adversarial Networks](https://mlsanity.com/paper/1812.04948/) (2018)
- [Chatbot Arena: An Open Platform for Evaluating LLMs by Human Preference](https://mlsanity.com/paper/2403.04132/) (2024)
- [FaceForensics++: Learning to Detect Manipulated Facial Images](https://mlsanity.com/paper/1901.08971/) (2019)
- [CNN-generated images are surprisingly easy to spot... for now](https://mlsanity.com/paper/1912.11035/) (2019)
- [Vision-Language Models for Vision Tasks: A Survey](https://mlsanity.com/paper/2304.00685/) (2023)
- [Exploring Plain Vision Transformer Backbones for Object Detection](https://mlsanity.com/paper/2203.16527/) (2022)
- [Towards Universal Fake Image Detectors that Generalize Across Generative Models](https://mlsanity.com/paper/2302.10174/) (2023)
- [Raising the Bar of AI-generated Image Detection with CLIP](https://mlsanity.com/paper/2312.00195/) (2023)
- [Online Detection of AI-Generated Images](https://mlsanity.com/paper/2310.15150/) (2023)
- [SynthID-Image: Image watermarking at internet scale](https://mlsanity.com/paper/2510.09263/) (2025)
- [More Real than Real: A Study on Human Visual Perception of Synthetic Faces](https://mlsanity.com/paper/2106.07226/) (2021)
- [A Representative Study on Human Detection of Artificially Generated Media Across Countries](https://mlsanity.com/paper/2312.05976/) (2023)
- [A Study of the Human Perception of Synthetic Faces](https://mlsanity.com/paper/2111.04230/) (2021)
- [Robots in the Middle: Evaluating LLMs in Dispute Resolution](https://mlsanity.com/paper/2410.07053/) (2024)
- [Methods and Trends in Detecting AI-Generated Images: A Comprehensive Review](https://mlsanity.com/paper/2502.15176/) (2025)
- [OpenFake: An Open Dataset and Platform Toward Real-World Deepfake Detection](https://mlsanity.com/paper/2509.09495/) (2025)
- [VER-Bench: Evaluating MLLMs on Reasoning with Fine-Grained Visual Evidence](https://mlsanity.com/paper/2508.04852/) (2025)
- [Analyzing Images of Legal Documents: Toward Multi-Modal LLMs for Access to Justice](https://mlsanity.com/paper/2412.15260/) (2024)
- [Can LLMs Create Legally Relevant Summaries and Analyses of Videos?](https://mlsanity.com/paper/2511.13772/) (2025)
- [LegalWebAgent: Empowering Access to Justice via LLM-Based Web Agents](https://mlsanity.com/paper/2512.04105/) (2025)

## Abstract (arXiv)

Visual evidence has long been treated as a reliable form of legal proof, but advances in artificial intelligence (AI) are undermining that assumption. This article asks how well humans and frontier multimodal large language models (MLLMs) can distinguish authentic evidentiary photographs from AI-generated counterparts in the object-centric scenarios typical of civil disputes. We built Synthetic Legal Evidence Detection (SLED-1400), a dataset of 200 authentic evidence images paired with 1,200 synthetic counterparts produced by six contemporary text-to-image generators across ten evidence categories. The same stimuli and response format were used in a controlled web experiment with 136 lay participants and in a standardized evaluation of four MLLMs (GPT-5.1, Gemini-3-Pro, Gemini-3-Flash, Qwen3-VL-235B). Human accuracy was 64.8% overall, and 48.5% and 51.0% on the two strongest generators (Gemini-3-Pro-Image and Flux-2-Max), indistinguishable from chance. MLLMs never misclassified an authentic image (100% specificity), but missed most synthetic outputs from the harder generators, with average MLLM detection at 5.9% on Gemini-3-Pro-Image outputs. Human and MLLM errors were largely uncorrelated, while the four MLLMs were strongly correlated with each other. Neither group is a reliable standalone authenticator. We argue that visual evidence in legal proceedings should be treated as inherently contestable, and that a workable procedural response must combine trained human review, MLLM screening, and provenance infrastructure such as C2PA Content Credentials.