AI Image Detector

Screens images using four independent AI detection models, including commercial engines like Sightengine GenAI, Winston AI, Was It AI, and an open-source cross-check (SAFE), to provide a more robust analysis.

Combines the outputs of multiple detection models using a published weighted-vote rule to deliver clear, graded conclusions (e.g., 'AI-generated', 'Real photograph') instead of ambiguous percentage scores.

Each report details the evidence, including the individual vote from each detection engine and visual inspection notes, ensuring transparency and allowing users to understand the verdict.

Prioritizes checking for C2PA credentials and generator self-tags. When these deterministic file evidences are present and certain, they automatically decide the verdict, overriding the AI model votes.

Allows users to upload and screen up to 50 images simultaneously, providing a triage summary and CSV export for efficient high-volume analysis.

Classifies against the fingerprints of current image models from major generators including Midjourney (v5–v7), DALL·E, Stable Diffusion, Flux, Google Imagen, Gemini, Adobe Firefly, and Ideogram.

Analyzes the exact bytes of uploaded images and records their SHA-256 hash, providing a verifiable chain-of-custody for forensic purposes.