# Google Hides the Sparkle — SynthID Stays When the Visible Mark Goes

Times of AI Desk · 2026-08-14 · Policy

[https://timesof.ai/2026/08/google-gemini-visible-watermark-toggle](https://timesof.ai/2026/08/google-gemini-visible-watermark-toggle)

> Google VP Josh Woodward said Gemini and Flow are rolling out a Media watermark toggle that removes the visible sparkle from Nano Banana, Omni, and Lyria outputs — while invisible SynthID and C2PA credentials remain. The toggle will not launch where law requires a visible mark. Same day as Anthropic’s Claude text-watermark explainer.

**Corrected October 14, 2026:** Corrected the toggle's status to rolling out, removed an unverified claim about OpenAI and Meta watermark practice, replaced a paraphrase that was presented as a quotation, and turned an internal reference into a link.

Brussels just turned on **transparency marking**. Google’s move splits the stack: **pretty outputs for users**, **forensic marks for platforms**. That is the opposite user-facing choice from Anthropic’s move to watermark Claude’s text.

**Google** said users can turn off the **visible “sparkle” watermark** on AI images, video, and music from **Nano Banana**, **Omni**, and **Lyria**. The control is a **Media watermark** setting in **Gemini** and **Flow**; **Search** is planned later. **Invisible SynthID** and **C2PA** metadata **stay on**. The toggle **will not launch in countries that require visible watermarks**. Same calendar day as Anthropic’s **Claude text-watermark** explainer; the two are separate announcements.

## What shipped

| Item | Detail (Josh Woodward / GeminiApp on X; The Verge, TechCrunch) |
|------|----------------------------------------------------------------|
| **Visible mark** | Optional; sparkle in the bottom-right corner can be hidden |
| **Invisible mark** | **SynthID** + **C2PA** still embedded |
| **Surfaces** | Gemini, Flow; Search “coming soon” |
| **Models named** | Nano Banana, Omni, Lyria |
| **Geo limit** | No toggle where law requires a visible mark |

Google’s VP for Gemini / Labs / AI Studio **Josh Woodward** posted the change; The Verge published a settings screenshot. Distinct from Anthropic Claude text watermark (same day): Claude is **invisible statistical marking of future text** for the EU AI Act. Google is **making a consumer-visible label optional** on **media**.

## Limits

- Whether EU/UK still treat SynthID-only as enough when the sparkle is off is unsettled.
- Search toggle not yet shipped.

## Sources

- [The Verge: You can now turn off Google Gemini’s visible watermarks (August 14, 2026)](https://www.theverge.com/tech/980416/google-gemini-ai-watermarks-removal)
- [TechCrunch: Google will now allow users to remove visible watermark from its AI generations (August 14, 2026)](https://techcrunch.com/2026/08/14/google-will-now-allow-users-to-remove-visible-watermark-from-its-ai-generations/)
- [Josh Woodward / GeminiApp posts on X (August 14, 2026)](https://x.com/joshwoodward/status/2088259244193960386)
- Times of AI: [Claude text watermark](https://timesof.ai/2026/08/anthropic-claude-text-watermark) (August 14, 2026)