AI Image Detection · 2026
Off-the-shelf classifiers give you a percentage. Clairfy gives you a verdict, a confidence bucket, and the specific visual signals that drove the call — in plain language.
Sample verdicts
Diffusion-model output. Skin texture is unnaturally smooth, lighting is internally inconsistent across the face, and the background bokeh repeats in soft, regular intervals.
Sensor noise, optical aberrations, and lighting are mutually consistent. No artifacts associated with diffusion sampling were detected.
Signals point in opposite directions. Photographic noise is present, but specific regions exhibit structural irregularities the model cannot resolve.
Three steps, three seconds
Drag, drop, or paste a URL. JPG, PNG, WebP, GIF — up to 5MB. Anonymous use works; sign in for more.
Frontier vision LLMs analyze pixel statistics, optical signatures, anatomy, lighting, and texture coherence — then commit to a verdict.
Not just AI / not-AI — a confidence bucket, a written rationale, and the specific visual signals that drove the call.