How can we tell if content was made by AI?

The answer

As AI-generated content floods the internet, two complementary standards are emerging to verify authenticity. Google DeepMind's SynthID embeds imperceptible statistical watermarks directly into AI-generated images, video, audio, and text during creation. Since its 2023 introduction, SynthID has watermarked over 20 billion pieces of content across Google's platforms, and the watermarks survive common transformations like resizing, cropping, and compression. Meanwhile, the Coalition for Content Provenance and Authenticity (C2PA) - backed by Adobe, Microsoft, Google, and others - attaches tamper-resistant metadata recording a file's creation tool, creator, and edit history. Google now combines both approaches: SynthID watermarks persist even when metadata is stripped, while C2PA credentials provide richer provenance when platforms preserve them. The challenge remains that C2PA metadata can be removed by screenshotting or re-uploading images.