Visual Tells in AI-Generated Images

Several artifacts appear consistently in AI-generated imagery in 2026. Teeth are often geometrically inconsistent, too symmetrical, too uniform, or slightly blurred. Ears frequently show deformation or asymmetry. Fine text in backgrounds becomes garbled. Hands and fingers are a persistent weakness of current models. Background reflections do not match foreground lighting. Hair at the hairline often blends unnaturally with skin or background.

Limitations of Visual Inspection

Visual inspection is unreliable for detecting high-quality deepfakes produced by current diffusion models. The artifacts described above are present in lower-quality outputs but may be absent or imperceptible in professionally processed deepfakes. Do not assume an image is real simply because you cannot identify visual artifacts.

Automated Detection Tools

Several AI-powered detection tools exist. Their accuracy varies significantly by model and output quality. No automated detector achieves 100% accuracy. For legal purposes, a detection result from an automated tool is supporting evidence, not conclusive proof. Professional forensic analysis remains the standard for litigation.

What to Do If You Find a Deepfake of Yourself

Document the URL, take screenshots with visible metadata, note the date of discovery, and do not contact the perpetrator directly. Immediately file a Section 223a notice with the hosting platform. Report to local law enforcement. File an FTC complaint if the platform fails to act. ScanErase provides biometric scanning to identify all instances across platforms and handles legal notice dispatch on your behalf.