- SynthID Detector is a free tool from Google that identifies invisible digital watermarks in multimedia files.
- It is compatible with content created by Google and technology partners such as OpenAI, NVIDIA and Kakao, with Apple planned for the future.
- Its detection is resistant to basic edits, although it is not a universal system since it only recognizes compatible brands.
Today, distinguishing between what is real and what has been manufactured by a machine is a real headache. With the massive deployment of generative artificial intelligence, the internet has become flooded with content that seems authentic but is pure algorithmic invention, forcing us to seek tools to help us avoid swallowing every hoax.
To address this, Google has decided to open its SynthID Detector to the general public , a web platform designed to detect invisible digital watermarks in multimedia files. What was once the domain of journalists and fact-checking experts is now available to anyone who wants to know if that viral video or curious photo has been filtered by AI.
What exactly is SynthID and how does it work?
Unlike standard metadata, which any user with minimal technical knowledge can delete, SynthID uses a technology developed by Google DeepMind that embeds imperceptible signals directly into the file's data structure. This means the watermark doesn't alter the visual or audio quality of the content, but remains as a digital signature.
The system is quite robust, as these signals typically withstand common modifications such as image cropping, the application of certain filters, or file compression when sharing on social media. However, it's important to understand that if a file undergoes extreme transformations or is repeatedly processed by various AIs, the watermark could be damaged and become undetectable.
Steps to use the synthetic content detector
Using the tool is a breeze and requires no complicated software installation, as it works directly from your browser at synthid.com . The process is very intuitive: once there, the user simply logs in and uploads the file they want to analyze, either by dragging it to the corresponding box or selecting it from their device.
To access the service, a Google, Apple, or ChatGPT account is required . Once the material is uploaded, the system scans the file for the digital signature and returns a result indicating whether or not the signature was detected. It's important to note that, for the time being, the platform's interface operates primarily in English.
Supported formats and scope of the tool
SynthID Detector's versatility is remarkable, as it doesn't limit itself to just images. In the video department , it can process files in MP4, MOV, and WEBM formats. When it comes to audio , the list is even broader, accepting WAV, MP3, OGG, FLAC, AAC, and M4A. As for images , it supports virtually everything: from classic JPG and PNG to more modern formats like WebP, AVIF, HEIC, HEIF, TIFF, and GIF.
A key point is that Google doesn't want to go it alone in this endeavor and is aiming to create an industry standard. Therefore, the tool not only detects Gemini content but also recognizes brands from OpenAI, NVIDIA, and Kakao . Furthermore, it has been confirmed that Apple will soon join this verification ecosystem, which will greatly expand the detector's usefulness.
Important limitations you should know about
This is where caution is needed to avoid jumping to conclusions. The SynthID Detector is not a universal AI detector. This means that if content was created with AI that doesn't use this watermarking system, the tool will report no watermark, but that doesn't mean the file is authentic.
In short, a negative result is not proof of authenticity, but simply the absence of a compatible signature . Furthermore, the system does not detail what specific changes were made to the file; it only confirms whether AI intervention occurred. In the case of videos, Gemini can pinpoint specific segments where the signature was detected, clarifying that the entire clip is not necessarily synthetic.
Other strategies for detecting deepfakes and fake content
Since no tool is infallible, it's advisable to combine the use of SynthID with a critical eye. One effective technique is to look for visual artifacts , such as unreadable text, deformed fingers, shadows that don't match the light, or patterns that appear overly repetitive and artificial.
It's also very useful to use Google's reverse image search to trace the true origin of a photo. Finally, analyzing the EXIF data of the original file can reveal the camera model used or the editing software employed, providing additional clues about the legitimacy of the material.
Google has integrated this technology on a massive scale, already marking more than 180.000 billion images and videos , as well as an amount of audio equivalent to 240.000 years. With the integration of these features into Chrome and Google Search , the fight against disinformation and deepfakes gains a powerful ally that processes more than a million requests daily, helping to restore a degree of security in the digital environment.