mlsanityData to 14 Oct 2026

Characterizing Photorealism and Artifacts in Diffusion Model-Generated Images

arXiv 2502.11989 ↗First posted 17 Feb 2025cs.HC cs.AI cs.CVcited by 14

On the map: Diffusion models → Image generation, text-to-image and editing

Abstract (arXiv)

Diffusion model-generated images can appear indistinguishable from authentic photographs, but these images often contain artifacts and implausibilities that reveal their AI-generated provenance. Given the challenge to public trust in media posed by photorealistic AI-generated images, we conducted a large-scale experiment measuring human detection accuracy on 450 diffusion-model generated images and 149 real images. Based on collecting 749,828 observations and 34,675 comments from 50,444 participants, we find that scene complexity of an image, artifact types within an image, display time of an image, and human curation of AI-generated images all play significant roles in how accurately people distinguish real from AI-generated images. Additionally, we propose a taxonomy characterizing artifacts often appearing in images generated by diffusion models. Our empirical observations and taxonomy offer nuanced insights into the capabilities and limitations of diffusion models to generate photorealistic images in 2024.

Key facts. Cited by 14 papers in the map (data through 14 Oct 2026). Placed in Image generation, text-to-image and editing, one of 14 directions in Diffusion models. Builds on 48 of its 90 references that are in the map. First posted on arXiv 17 Feb 2025.

Builds on

48 of 90 references.

Closest ancestors · 5

Same direction, newest first

  1. 2024How to Distinguish AI-Generated Images from Authentic Photographs12 Jun
  2. 2024Detecting Multimedia Generated by Large AI Models: A Survey22 Jan
  3. 2023Exposing the Fake: Effective Diffusion-Generated Images Detection12 Jul
  4. 2023Intriguing properties of synthetic images: from generative adversarial networks to diffusion models13 Apr
  5. 2022Towards the Detection of Diffusion Model Deepfakes26 Oct

Other references · 43

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  2. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,380
  3. 2018A Style-Based Generator Architecture for Generative Adversarial Networkscited by 6,013
  4. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 4,356
  5. 2019Analyzing and Improving the Image Quality of StyleGANcited by 3,588
  6. 2017Progressive Growing of GANs for Improved Quality, Stability, and Variationcited by 2,988
  7. 2018Large Scale GAN Training for High Fidelity Natural Image Synthesiscited by 2,626
  8. 2015Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networkscited by 2,054
  9. 2022Diffusion Models: A Comprehensive Survey of Methods and Applicationscited by 1,114
  10. 2019CNN-generated images are surprisingly easy to spot... for nowcited by 847
  11. 2020Leveraging Frequency Analysis for Deep Fake Image Recognitioncited by 504
  12. 2023Towards Universal Fake Image Detectors that Generalize Across Generative Modelscited by 490
31 more
  1. 2018Exposing Deep Fakes Using Inconsistent Head Posescited by 297
  2. 2019Detecting and Simulating Artifacts in GAN Fake Imagescited by 293
  3. 2020Watch your Up-Convolution: CNN Based Generative Deep Neural Networks are Failing to Reproduce Spectral Distributionscited by 280
  4. 2020The Creation and Detection of Deepfakes: A Surveycited by 247
  5. 2023Stable Bias: Analyzing Societal Representations in Diffusion Modelscited by 226
  6. 2018Do GANs leave artificial fingerprints?cited by 190
  7. 2023Raising the Bar of AI-generated Image Detection with CLIPcited by 164
  8. 2022Human heuristics for AI-generated language are flawedcited by 105
  9. 2023Art and the science of generative AI: A deeper divecited by 95
  10. 2024Can ChatGPT Detect DeepFakes? A Study of Using Multimodal Large Language Models for Media Forensicscited by 69
  11. 2023Shadows Don't Lie and Lines Can't Bend! Generative Models don't know Projective Geometry...for nowcited by 61
  12. 2019HYPE: A Benchmark for Human eYe Perceptual Evaluation of Generative Modelscited by 59
  13. 2020Exposing GAN-generated Faces Using Inconsistent Corneal Specular Highlightscited by 59
  14. 2021Deepfake Detection by Human Crowds, Machines, and Machine-informed Crowdscited by 58
  15. 2021Eyes Tell All: Irregular Pupil Shapes Reveal GAN-generated Facescited by 48
  16. 2022Lighting (In)consistency of Paint by Textcited by 35
  17. 2022Perspective (In)consistency of Paint by Textcited by 34
  18. 2020Detecting CNN-Generated Facial Images in Real-World Scenarioscited by 32
  19. 2023Qualitative Failures of Image Generation Models and Their Application in Detecting Deepfakescited by 31
  20. 2019Exposing GAN-synthesized Faces Using Landmark Locationscited by 30
  21. 2021More Real than Real: A Study on Human Visual Perception of Synthetic Facescited by 23
  22. 2023Context Understanding in Computer Vision: A Surveycited by 23
  23. 2023AI-Generated Image Detection using a Cross-Attention Enhanced Dual-Stream Networkcited by 22
  24. 2024Organic or Diffused: Can We Distinguish Human Art from AI-generated Images?cited by 21
  25. 2023Detecting Generated Images by Real Images Onlycited by 20
  26. 2024RLVF: Learning from Verbal Feedback without Overgeneralizationcited by 19
  27. 2022Human Detection of Political Speech Deepfakes across Transcripts, Audio, and Videocited by 16
  28. 2024People are poorly equipped to detect AI-powered voice clonescited by 13
  29. 2023Finding AI-Generated Faces in the Wildcited by 8
  30. 2024Crafting Synthetic Realities: Examining Visual Realism and Misinformation Potential of Photorealistic AI-Generated Imagescited by 6
  31. 2022Identifying the Context Shift between Test Benchmarks and Production Datacited by 5

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 4; 2026: 10.

14 papers.

Sort:

By area:

  1. 2025AIGI-Holmes: Towards Explainable and Generalizable AI-Generated Image Detection via Multimodal Large Language Modelscited by 33
  2. 2025DiffusionReward: Enhancing Blind Face Restoration through Reward Feedback Learningcited by 5
  3. 2025GeneVA: A Dataset of Human Annotations for Generative Text to Video Artifactscited by 2
  4. 2025Review of Hallucination Understanding in Large Language and Vision Modelscited by 2
  5. 2026Human-AI Ensembles Improve Deepfake Detection in Low-to-Medium Quality Videoscited by 1
  6. 2026SalArt-VQA: Diagnosing Whether VLMs Understand Salient Artifacts in Generated Imagescited by 1
  7. 2026Illusion or Integrity? Geometrical Consistency Metric for AIGC Video Quality Evaluationcited by 1
  8. 2026Na"ive PAINE: Lightweight Text-to-Image Generation Improvement with Prompt Evaluationcited by 0
  9. 2026Chroma Clues: Leveraging Color Statistics to Detect Synthetic Imagescited by 0
  10. 2026Generative AI Literacy Training Improves Intelligence Analysts' Discrimination of Real and AI-Generated Imagescited by 0
  11. 2026Hidden-Shot: Towards One-Shot Task Generalization for Low-Level Vision Generalist Modelscited by 0
  12. 2026Robustness of AI-Art Detectors under Generator Shiftcited by 0
2 more
  1. 2026Steering Fields: Adaptive Vector Fields for Safe Image Generation and Beyondcited by 0
  2. 2026TRIG: Toward Measuring Texture Realism in Image Generationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Synthetic Photography Detection: A Visual Guidance for Identifying Synthetic Images Created by AI
  2. 2025Detecting AI-Generated Images via Diffusion Snap-Back Reconstruction: A Forensic Approach
  3. 2023On quantifying and improving realism of images generated with diffusion
  4. 2024The Cat and Mouse Game: The Ongoing Arms Race Between Diffusion Models and Detection Methods
  5. 2024Unveiling the Truth: Exploring Human Gaze Patterns in Fake Images
  6. 2024Detecting Origin Attribution for Text-to-Image Diffusion Models
  7. 2024Image-Perfect Imperfections: Safety, Bias, and Authenticity in the Shadow of Text-To-Image Model Evolution
  8. 2026Repurposing Image Diffusion Models for Adversarial Synthetic Structured Data: A Case Study of Ground Truth Drift