Google’s Launches its Own AI Checker - SynthID for AI images, text and more

Google has unveiled its new SynthID tool to watermark text generated by AI, after previously embedding invisible watermarks in AI-generated images and audio. Developed by Google DeepMind, SynthID now supports watermarking and detecting machine-generated text across Vertex AI and its open model library.

This move targets misinformation and content manipulation. By embedding signals directly into text and images without disrupting their appearance or meaning, SynthID helps flag synthetic content while keeping it readable.

Also read: Google I/O 2025: Top 10 Highlights of Google’s New Announcements

Unlike standard metadata or visible labels, SynthID’s watermarks remain even after editing or copying. DeepMind claims its detection model performs reliably even when the content is paraphrased. The system relies on subtle, encoded patterns, making it difficult to strip or bypass the tag.

Initially, Google deployed SynthID for images created with Imagen. In November 2023, it extended support to AI audio from YouTube’s Dream Track and tools like Lyria.

Google is pitching SynthID as a technical answer to rising concerns over the misuse of generative AI, especially in election cycles and media. While not foolproof, it gives developers and platforms a new way to verify content origins.