About OriginVerdict
OriginVerdict is an independent image-origin review tool built around a simple principle: a detector should show uncertainty instead of hiding it.
Generative images are improving quickly, while real photographs are routinely compressed, edited, screenshotted, and stripped of metadata. In that environment, a confident-looking percentage can be useful, but it can also be misunderstood. OriginVerdict pairs the visual model signal with limited file clues, a visible uncertainty band, and plain-language failure modes.
What we are building
The first OriginVerdict tool is an AI photo detector for JPG, PNG, and WebP images. The product direction is broader than one page: future origin-review tools may cover video, audio, provenance credentials, or other media when they can be implemented and explained responsibly.
Product principles
- Evidence before certainty. Results are framed as signals, not proof of authorship or deception.
- Visible limitations. Known failure modes belong beside the result, not in fine print.
- Minimal image retention. Originals use private temporary storage, are deleted after the synchronous check, and have a one-day automatic cleanup fallback.
- Specific claims. We avoid publishing unsupported accuracy numbers or implying that one model can detect every generator.
- Useful context. Safe metadata clues are shown when available, while GPS parsing is disabled.
Independence and providers
OriginVerdict uses third-party infrastructure and currently uses Sightengine for the pixel-based AI-generation score. OriginVerdict is not affiliated with, endorsed by, or sponsored by Sightengine or the creators of image-generation products mentioned on the site.
Contact
For support, privacy requests, partnerships, or reproducible detector feedback, email support@originverdict.com.