A new iPhone app called Verdict aims to identify AI-generated images and videos on-device, offering privacy benefits but variable accuracy in tests, highlighting ongoing challenges in digital content authentication.
Verdict is a new iPhone app that aims to help users decide whether a photo or video was made by artificial intelligence. According to the developer’s description in the App Store, it offers a 0–100 score and a plain-language assessment, while keeping processing on the handset rather than sending files to the cloud. It requires iOS 16 or later, works without an account and does not upload the media being checked.
The app’s approach is deliberately forensic rather than purely statistical. For still images, Verdict combines an AI-detection model with metadata inspection and sensor-noise analysis. Its breakdown also includes a compression-related check, but the developer says that signal is shown for context only and does not affect the final score. For video, the app samples 15 frames across the clip, scores each one, and uses the median result, which is intended to reduce the effect of outliers. The developer says scan history remains on the device and that Apple’s photo picker limits access to the selected file rather than the full library.
Testing by Help Net Security suggests the app can produce plausible results on some content, but also some sharply conflicting ones. A garden image was judged 51% real, another image was labelled 72% AI-generated, and a photograph of a house was marked fully real. The results screen exposed the contribution of each check, including instances where they disagreed, which is useful for judging confidence rather than treating the score as absolute.
The main weakness appeared in a concert video, which Verdict classified as 100% AI-generated even though the same screen identified it as camera-captured footage. That kind of error matters because it shows how lighting, motion, compression or other processing can confuse automated detection. Other mobile tools in the same space, including FrameProof, FakeOut, Verifyco and Pixoraid, also position themselves around evidence-based screening, combining multiple signals or keeping uncertainty visible rather than pretending to deliver perfect certainty. In Verdict’s case, the offline design and on-device privacy are clear strengths, but its output still needs to be weighed against context, source and common sense.
Disclaimer: This content is intended for informational purposes only. Readers are advised to exercise their own judgement, conduct due diligence, or consult a qualified expert before acting on any information provided.
