Meta Builds AI Detection Tool to Identify Images and Video Created With Its New Models

Wed, 08 Jul 2026 00:00:00 +0000

Meta has unveiled a new web-based detection tool to identify images and videos created with its latest AI models. The tool detects invisible watermarks embedded through its proprietary “Content Seal” system.

Meta’s new Content Seal watermarking system aims to provide a reliable way to identify AI-generated images and videos. (Image: geralt / Pixabay)

What Is Content Seal?

Content Seal is Meta’s invisible watermarking technology that remains embedded in media “even when cropped, compressed, resized, or screenshotted.” Unlike earlier versions of Meta AI that placed a small visible logo on generated images, the new system is entirely invisible to the naked eye.

The technology is currently integrated with Muse Image, Meta’s latest image generation model. While the version shipping with Muse Image is proprietary, Meta has previously released open-source versions of similar watermarking technology.

How the Detection Tool Works

The web-based interface lets users upload images to check for a Content Seal watermark. A positive result means the image was generated or edited using Meta AI, while a negative result means it’s “unlikely” the image was processed with Meta’s tools. Early tests by Engadget confirmed the tool successfully detects watermarks on AI-generated images, AI-edited photographs, and even screenshots of AI content.

Limitations and Gaps

The tool currently only works with Muse Image — older Meta AI models are not supported. Content Seal is also incompatible with SynthID or C2PA Content Credentials, two established watermarking methods used elsewhere. Additionally, the tool enforces a daily rate limit on checks, limiting its practicality for widespread use.

Oversight Board Concerns

Meta has faced criticism over its inconsistent AI labeling. The Oversight Board earlier this year flagged concerns that Meta was “inconsistently implementing” digital watermarks on AI content. Notably, the detection feature has not been integrated into the Meta AI app itself — when asked about an image the web tool had already flagged, Meta’s app assistant said it lacked the ability to detect AI-generated content.

What’s Next