AI Image Detection: Expert Tricks to Spot Fakes 2026

Detecting AI-generated images has become a cat-and-mouse game, and experts say it will only get harder. As models become more precise, the task is far more complex than it was just a few years ago.

Not long ago, we could spot AI images by looking at hands and counting fingers. Now, with increasingly sophisticated image and video models, that's no longer enough.

The good news is that there are still ways to detect if an image has been generated by AI, though the pace of advancement means this could change soon. Detection is less intuitive than before, but paying attention to geometry, shadows, and perspective—essentially technical drawing—can help.

The Expert Behind the Tricks

Hany Farid, a specialist at the University of California, Berkeley, and one of the world's leading experts in image forensics, has spent over two decades determining whether a photo or video has been manipulated.

Santiago Lyon, former director of photography at Associated Press and now working in digital security at Adobe, describes Farid in a Science report as "a kind of dean of digital forensics" precisely because he has been in the field for so long. Farid helped found...

Why AI Images Are Getting Harder to Spot

The rapid improvement of generative models means that the telltale signs of AI generation are fading. Early diffusion models often produced obvious errors like extra fingers or distorted facial features.

Today's models have learned from those mistakes and produce images that are nearly indistinguishable from real photographs to the untrained eye. This makes traditional visual inspection less reliable.

Expert Tricks for Detecting AI Images

According to Farid and other experts, the key is to focus on physical consistency. AI models often struggle with the laws of physics, especially when it comes to lighting and geometry.

• Check shadows and reflections: AI often messes up how shadows fall or how reflections appear on surfaces.
• Examine perspective: Lines that should converge may not, or objects may appear at impossible angles.
• Look for geometric inconsistencies: Straight lines might be slightly curved, or patterns may not align properly.
• Analyze fine details: Zoom in on textures like skin, hair, or fabric—AI sometimes produces unnatural smoothness or repetitive patterns.
• Consider the context: Does the lighting match the environment? Are there any anachronisms or out-of-place elements?

Tools and Techniques Used by Forensics Experts

Beyond visual inspection, forensic analysts use specialized software to detect manipulation. These tools can analyze metadata, compression artifacts, and noise patterns that are invisible to the human eye.

However, as AI models improve, these tools also need to evolve. Farid's work focuses on developing new methods to keep up with the advancements. The background to this story is set out in ISO 12312-2 Eclipse Glasses: How to Spot Safe vs. Dangerous Fakes. A related thread runs through Tricks to Paint like an Expert!. Earlier coverage in The best tricks to use your tools like an expert! shows how the argument developed.

The Future of AI Detection

The arms race between AI generation and detection is ongoing. As models become more sophisticated, detection methods must also advance. A closer look at the same subject appears in Complete guide with 10 tricks that every carpentry expert should know!. Anyone following the wider picture can start with AI Images Threaten Elections as Deepfakes Spread Rapidly. The same theme surfaces in MEP Fights Nude Deepfakes as EU Urged to Act on AI Abuse, which takes a different angle.

Experts like Farid warn that we may reach a point where it becomes extremely difficult to distinguish AI-generated content from real content without specialized tools. This raises concerns about misinformation and trust in visual media. Context on how this played out before sits in Meta Ordered to Remove Deepfakes: AI Video Safeguards Inadequate. A companion piece, AI Deepfake Expert Hany Farid Says Visual Detection Is Nearly Useless, approaches the same subject from another angle.

FAQ