Nowadays, AI-generated images and digital content are increasing rapidly. Many people are no longer able to identify whether a photo, graphic or image has been created by AI or is real. To solve this problem, Google has introduced a new feature in its Gemini app. This is called SynthID detection. This feature confirms for you whether a photo has been created with a Google AI model or not. Now users can easily know whether the content is original or AI-generated by uploading images directly in the app.
Google has added a new feature to its Gemini app, which helps in finding out whether an image has been created by AI or not. This is called SynthID detection. Through this feature, users can now easily check whether a photo has been created by Google AI model or not.
How does Gemini 3.0 and SynthID work?
Google AI models add an invisible watermark called SynthID to the images they generate. This watermark is not visible to the human eye, but it can be decoded through Google’s special tools. Currently, SynthID detection only works for Google AI models.
Earlier this facility was available only for journalists and media professionals in SynthID Detector Portal. Now it has been added to the Gemini app, so every user can recognize the AI-generated image for themselves. Information about AI attribution and SynthID watermark can also be seen through the ‘AI Info’ panel in Google Photos.
How to check if an image is AI-generated or not?
- Open the Gemini app or Gemini web version.
- Upload your image.
- Ask the question: ‘Was this image generated by AI?’
- Gemini will analyze the image and tell you if the SynthID watermark is present.
Even if a watermark is not found, Gemini looks for other signs of AI in the image. As:
- Prompt excerpts in text
- Anomalous state of physics or objects
- Too smooth or unnatural ‘AI skin’
- File names and watermarks of some AI platforms
From these signs it can be determined that the image has been created by AI, but it does not give 100% certainty.
Limitations of SynthID
The biggest limitation is that SynthID only works for Google AI models. Other platforms like Midjourney or GPT Image 1 are not included.
- Gemini can look for other signals in the images from these platforms, but cannot provide definitive information.
- Cropping, resizing or applying filters does not affect the watermark, but technically knowledgeable people can easily remove it.
This shows that technical protections in AI detection are not completely sufficient and we need to be vigilant in identifying AI-related content.
More features in future
Google has promised that the SynthID feature will also work in video, audio and text. It will be available to more users in upcoming updates to the Gemini app.
SynthID detection is a small, but important step. This helps users learn not to trust an image just by looking at it. AI and Deepfake content are on the rise, so Gemini’s new feature will help people get into the habit of being vigilant and checking AI content. However, this feature is not fully available in all countries yet but it is expected that it will be rolled out gradually.
