SynthID Detector, Google’s new technology for identifying AI-generated content
Share
The proliferation of AI verification poses an urgent challenge: how do we know if what we see is real or has been created by a machine? Against this backdrop, Google has unveiled SynthID Detector, a solution that identifies AI-generated content.
In 2023, Google introduced SynthID, a deep learning-based watermarking technology that allows AI-generated images and sounds to carry a watermark that is detectable but imperceptible to humans. Now, two years later, SynthID Detector has been launched, a Google DeepMind initiative designed to tackle deepfakes, fake videos, and other forms of digital manipulation by detecting these AI-generated manipulations.
How does the SynthID Detector work?
The SynthID Detector combines two key processes that reflect the sophistication of Google AI tools:
- Watermarking AI-generated content: When models such as image3 or Lyria are used, SynthID embeds a watermark in the content without altering its appearance or sound.
- AI-generated content detection: SynthID Detector can then detect the watermark, even if the content has been cropped, compressed, or altered, to confirm its AI origin.
This level of watermark detection is crucial in an era where Ai-generated videos, AI-generated images, and Ai-generated audio are nearly indistinguishable from real content.
Benefits of SynthID Detector in combating misinformation and fake news
In a world where fake videos go viral in minutes, tools like SynthID Detector represent a critical barrier against mass deception. Here are some key advantages:
- Invisible and persistent marks that survive basic modifications.
- Confidence for the end user: helps to differentiate between authentic and synthetic.
- Integration potential with systems like the Gemini model or Google Veo.
In short, SynthID Detector sets a new standard in generative AI content regulation.
What are the limitations of SynthID Detector?
Despite its innovation in the AI verification, SynthID Detector comes with some limitations:
- It only detects content that was watermarked by SynthID —unmarked files remain undetected.
- Extreme edits can diminish the detectability of the watermark.
- Currently focused on images and audio; text watermarking and full video watermarking are still evolving.
This creates the need for complementary tools—solutions that can detect AI content regardless of its watermark status.
VerifAI or SynthID Detector?
At TU, we developed VerifAI with a clear mission: to detect audio, image, or video content generated or manipulated with artificial intelligence.
Unlike SynthID Detector, VerifAI acts as an authenticity test. It uses multiple analysis engines to examine images, videos, or audio, identifying typical AI-generated patterns—even if the content has been edited or its origin is unknown.
Both solutions are complementary, offering a comprehensive defense against digital fraud and misinformation.
Why you should integrate VerifAI alongside Google’s SynthID
If your workflow involves reviewing or distributing digital media, combining SynthID Detector and VerifAI gives you a comprehensive defense:
- SynthID Detector detects whether content has SynthID watermarks.
- VerifAI verifies audio, video, or image content generated or manipulated by AI in a wide variety of contexts. It is useful for media outlets that deal with fake videos, hoaxes, or misinformation on a daily basis, as well as for sectors that require identity verification (KYC), insurance, technology platforms, and legal, financial, or educational environments, where content authenticity is critical.
How to get started with SynthID and VerifAI
SynthID Detector is currently embedded into AI models like Imagen, Lyria, and Gemini. Future plans include integration with tools such as Google Veo.
Meanwhile, VerifAI is already live and can be tested via TU’s website, where users can assess content to check if it was generated or manipulated by AI.
As the digital world becomes increasingly shaped by generative AI, the ability to distinguish real from artificial is becoming essential. Tools like SynthID and VerifAI are leading this transformation, offering layered protection against AI deception, fraud, and misinformation.
What is SynthID and SynthID Detector by Google?
SynthID is a tool developed by Google deepmind that inserts an invisible watermark in AI verification, such as images and audio, to facilitate their identification and prevent digital tampering. SynthID Detector checks whether the watermark is present or not.
How can I detect if content was generated by AI?
Detection can be done using tools like SynthID Detector if the content was previously marked, or with solutions like VerifAI, which can identify AI-generated content without requiring a watermark, using deep learning and various analysis engines.
Does VerifAI require content to be watermarked?
No. VerifAI can analyze any image, text, or audio file to determine if it was created by AI, even if it lacks a watermark.
What’s the difference between SynthID Detector and VerifAI?
SynthID embeds the watermark when content is generated by AI. SynthID Detector checks for the presence of that watermark. VerifAI analyzes existing content to detect whether it has been artificially generated or digitally manipulated. They are complementary tools.
Can SynthID Detector be used with video?
For now, SynthID Detector is focused on image and audio formats. Google is working to extend it to AI-generated video. Meanwhile, other tools can help analyze fake videos and deepfakes.
Where can I try VerifAI?
You can access VerifAI here to start verifying AI content with advanced detection tools.
SEO, ASO and CRO Specialist at Telefónica Innovación Digital | Growth, Strategy and Conversion Optimization.
