Google Allows Removal of Watermarks from AI Images
Starting August 2026, users can remove visible labels from generated content, but invisible SynthID and C2PA metadata remain mandatory.
Starting August 2026, users can remove visible labels from generated content, but invisible SynthID and C2PA metadata remain mandatory.
Starting in August 2026, Google has provided users with the ability to remove visible watermarks from AI-generated content, making it harder to distinguish authentic photos from those created by artificial intelligence. Josh Woodward, Google's vice president, confirmed the news on social network X, emphasizing that the company seeks a balance between creative freedom and safety.
Previously, generated images had a visible label in the bottom right corner, which bothered some individuals because ChatGPT images do not have such a label. Thanks to the new setting, users will be able to choose whether to keep the visible watermark, changing this situation. This option will be available in Gemini and Flow, with plans to extend it to AI Mode within Google Search.
The change will not apply only to images. It will also be possible to remove visible watermarks from videos created with Gemini Omni and music generated by Lyria. The only exception will be countries where retaining such labels is legally required.
However, removing the visible watermark will not leave AI content completely unmarked. Google plans to retain the use of invisible SynthID labels and C2PA metadata. As a result, users will be able to ask Gemini whether a specific image was generated by artificial intelligence. If SynthID is present on the image, Gemini will be able to recognize that it is AI-generated content.
Just a few years ago, recognizing artificially generated images was simple. Extra fingers, deformed faces, distorted text-these errors were visible at first glance. Today's generators, such as Gemini, Nano Banana, and Omni models, are so sophisticated that it is practically impossible to distinguish them from authentic photos.
The problem is not just technical. Deepfake technology is already being used in scams where victims fall for fake videos and lose control of their profiles. Although legal frameworks are increasingly penalizing such abuses, the ability to remove visible labels opens the door to much subtler manipulations that may not necessarily end up in court.
Google defends itself with the argument that invisible watermarks and metadata remain in place and that anyone can verify the authenticity of an image through Gemini tools or Google Search. However, the question arises that the Mountain View company avoids: how many average users, without technical knowledge, will even be aware that they need to check the source of an image before trusting it or sharing it?
Recall that Google introduced the first watermarks for AI-generated images back in 2023, and Google Photos later also labeled photos edited with AI. This current decision represents a step backward in transparency, regardless of technical reasons. The visible label had a preventive role, acting before someone would start checking.
In an era when misinformation spreads faster than ever, this issue is not just technical but also social. Google's decision to make visible labels optional shifts the responsibility of verification onto the individual, and it is precisely this that creates room for manipulation. Instead of striving for greater transparency, the tech giant chooses flexibility, and public trust will be what pays the price.
