Your brand photos just got harder to spot — and easier to trust
You’ve probably seen it: the tiny sparkle icon in the bottom-right corner of a Gemini-generated image. If you’ve been using AI to create product shots, social media headers, or client mock-ups, that little mark has been a dead giveaway that your content was made by a machine. Earlier this week, Google quietly changed that. You can now toggle off the visible “Media watermark” in Gemini and Flow, Google’s AI video generator.
But before you rush off to flip that switch, understand what’s really happening. The sparkle you could see is optional now — but the watermark you can’t see stays. And for a Malaysian SME owner who values both speed and trust, this matters more than you might expect.
TL;DR
Google now lets you remove the visible “sparkle” watermark from AI-generated images, videos, and music. Invisible SynthID and C2PA tags remain embedded, so you can still verify what’s AI. For your business, this means cleaner visuals — but you need to keep a clear audit trail and stay honest with customers.
What this means
Let’s break it down simply. A visible watermark is the obvious label you see on an image — like Google’s sparkle icon in the corner. An invisible watermark is a tiny digital signature baked into the file itself. It can’t be seen by the naked eye, but software can read it. Google uses two types: SynthID, a deep-learning-based tag embedded in the pixels, and C2PA, a metadata record attached to the file. According to Josh Woodward, Google’s VP of Labs, Gemini, and AI Studio, both remain in place even after you switch off the visible marker.
How do you check? You can ask Gemini or Google Search whether a piece of content is AI-generated. This is a significant shift: you lose the eyesore, but you don’t lose accountability.
Visible watermarks were never a real security feature — they were a design compromise. The actual proof of AI involvement is the invisible metadata, and that’s staying.
How this applies to Malaysian SMEs
Your startup might use AI to mock up product packaging. Your restaurant might use AI to create a crispy photo of a new burger before the chef finishes testing the recipe. With the visible watermark gone, those visuals look polished enough to post straight to Instagram — and your customers won’t be able to tell the difference at a glance. That’s exactly why you need your own checks in place.
Consider this scenario: you run a small e-commerce store on Shopee or Lazada. You generate a lifestyle image of someone using your product. Without a visible label, a shopper might genuinely believe it’s a real customer photo. If the actual product doesn’t match, you’re looking at a refund request or a bad review. The invisible watermark won’t help your customer — only you can set the right expectations. For F&B, beauty, or property businesses, the stakes are even higher. Malaysian advertising regulations are strict about false claims, and misleading product photos can land you in hot water. The hidden metadata won’t save you from that.
On the marketing side, if you outsource content creation to a freelancer or agency, you need to know whether that glossy brochure image was generated by Gemini and had its watermark removed. When a client later asks, “did you create this from scratch?” you want to be able to prove it. The C2PA metadata is actually your friend here — it gives you a digital chain of custody for your assets. Keep your AI prompts and timestamps filed properly, and you’ll never end up in an awkward honesty conversation.
There’s also an export angle. If you sell digital products or design services to clients in Europe, remember that the EU’s AI labelling rules are now in effect. The Verge’s coverage notes that Google won’t enable the watermark toggle in countries that require visible watermarks. So if you operate or sell into those markets, keep the visible label on — or expect compliance checks from your overseas business partners.
What changed in Gemini’s watermarking
| Element | Before | After |
|---|---|---|
| Visible sparkle | Always present | Can be disabled |
| SynthID invisible tag | Embedded | Still embedded |
| C2PA metadata | Embedded | Still embedded |
| Verification method | Spot by eye | Ask Gemini or Google Search |
Practical takeaways for your business
- Start an AI asset log. For every AI-generated image, video, or music clip, record the tool, the prompt, and the date. A simple spreadsheet works.
- Verify before publishing. Ask Gemini or Google Search “is this AI-generated?” so you know what your customers and competitors can also find out.
- Match your market’s rules. If you sell to the EU or other regions with labelling laws, keep visible watermarks on.
- Disclose where it matters. If a customer might reasonably assume a photo is real — like a product shot or a client testimonial — say it’s AI-assisted.
- Lean on C2PA for proof. Use the embedded metadata to defend yourself if a supplier or client questions the origin of your content.
- Train one staff member. Make sure at least one person in your company knows how to inspect SynthID and C2PA tags using free tools.
The bigger picture
This update is part of a larger movement across the AI industry. OpenAI never used visible watermarks; it relies on SynthID and C2PA. Meta launched its own Content Seal standard. Anthropic recently announced invisible watermarks for text. The direction is clear: visible labels are fading, and invisible verification is becoming the norm.
For Malaysian SME owners, this means AI-generated content is about to become indistinguishable from human-made marketing material. The businesses that will come out ahead are the ones that embrace transparency — not because they’re forced to, but because trust is the only durable advantage left. When your competitor can generate a perfect product photo in ten seconds, your real edge is your actual product quality, your truthful descriptions, and the way you treat customers. Use AI to move faster, but invest in the things AI can’t fake: your workshop, your people, your consistency.
The sparkle may be gone, but the responsibility is now on you. Use the invisible markers to build a cleaner, more trustworthy business — not a more deceptive one.
Ready to Streamline Your Operations?
Your business should run itself. AutoRunBiz deploys AI agents to automate your daily operations — WhatsApp orders, invoicing, customer follow-ups, and accounting. Book a free 15-min ops audit to see where automation fits your business →