Deciding what’s real on the Internet won’t be easy in the future. The scale of AI‑generated media can be hard to grasp. Starling Lab estimates it took until 1975—149 years after the camera—for 1.5 billion images. Generative AI did it in 18 months. This spring, Google announced that its tools had created more than 100 billion AI images and videos in just a few years.

Google partnered to expand its Synth ID watermarking technology. The watermark is embedded in pixels and survives edits. OpenAI, Runway, Nvidia, and others are slated to use it.

Testing with Nano Banana Pro images showed Synth ID survived 300 compression iterations; only a 20‑percent crop broke detection. A 50‑percent crop broke it earlier, around 250 iterations.

Google limits verifier access to ~10 checks/day and no public API, making bypass harder. While not a silver bullet and hacks could emerge, the widespread adoption across consumer AI firms points to a more traceable, though fragmented, ecosystem.