Google will now allow users to remove visible watermarks from AI content
Google plans to let users remove the visible AI watermark from images and videos generated by several of its models, while retaining less obvious provenance markers. The change matters because visible labels are designed to help people quickly recognise AI-generated material, but Google says it is trying to balance that safety benefit with creative control for users.
The option will cover Nano Banana, Omni and Lyria initially, with Gemini and Flow to follow, but will not be available where visible watermarking is legally required, including the EU. Google says invisible SynthID watermarks and C2PA metadata will remain embedded, although checking them requires additional tools or actions; South Korea has also introduced comparable AI-labeling requirements.
- Google will let users hide visible AI watermarks in many regions.
- Invisible SynthID and C2PA markers will remain.
- EU legal requirements will prevent the toggle there.
New here? Start with this
Google’s AI tools can create images, video and audio from written prompts. Companies have increasingly added watermarks or labels to this material so that viewers can tell when something was made or altered by artificial intelligence.
Google is one of several major firms using both visible labels and hidden markers. Its SynthID system places an invisible signal in AI-made files, while C2PA metadata can record information about how content was created or edited.
The issue has become more important as realistic AI material spreads online, where it can be hard to distinguish from ordinary photographs, footage or recordings. Different countries are taking different approaches: some require clear visible labelling, while others rely more on voluntary company policies and technical tracing tools.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Supporters argue that removing an on-image label gives legitimate creators more control over the presentation of work they have generated or substantially developed with AI. They contend that persistent invisible SynthID and C2PA markers preserve traceability without imposing an always-visible badge that may be aesthetically intrusive, commercially limiting or misleading about the degree of human creative input. Keeping visible marks where law requires them is presented as a proportionate way to respect democratic rules while allowing flexibility elsewhere.
The case against
Critics argue that visible watermarks are valuable precisely because they let ordinary viewers identify AI-generated material immediately, without specialised tools or technical knowledge. If the remaining provenance signals are difficult to check, they may offer limited practical protection against deception, impersonation and the spread of manipulated media, especially on fast-moving social platforms. From this perspective, creative convenience should not weaken clear, universal disclosure at a time when public trust in images, video and audio is already fragile.