The scale of AI-generated media is staggering. Starling Lab, a research collaboration from Stanford University and the University of Southern California, estimates that it took until 1975 — 149 years after the invention of the camera — for humanity to create 1.5 billion images. Generative AI did the same in just 18 months. And AI didn’t stop there.
This spring, Google announced at I/O that its tools had been used to create more than 100 billion AI images and videos in just a couple of years. Google is far from the only source of AI content. The company coupled this staggering statistic with a slew of partnerships to expand the use of its SynthID watermarking technology, which can label AI-generated content and potentially help people identify what’s real.
Why SynthID is considered resilient
SynthID works by embedding an invisible digital watermark directly into the pixels of AI-generated images and videos. Google says it is designed to survive common edits like cropping, resizing, and compression — making it harder for bad actors to strip it off.
Early testing suggests the watermark is indeed robust against many tampering attempts. This is a meaningful technical achievement, especially as deepfakes and synthetic media become harder to distinguish from reality.
The core limitation: Watermarking is not a cure
Despite its resilience, SynthID does not solve the AI misinformation problem. The watermark only labels content after it is created. It does not prevent someone from generating misleading content in the first place, nor does it stop them from sharing it without the watermark if they use a different tool.
“Watermarking is a useful layer, but it’s not a silver bullet,” experts have noted. The technology relies on voluntary adoption by platforms and creators. If a bad actor uses an AI tool that does not embed SynthID — or removes the watermark before sharing — the system is powerless.
How the AI content explosion outpaces detection
The timeline is sobering. Starling Lab’s research highlights that the camera took 149 years to produce 1.5 billion images. Generative AI did it in 18 months. Google alone has generated over 100 billion AI images and videos in just a couple of years.
This exponential growth means detection tools are always playing catch-up. Even if SynthID were universally adopted, the sheer volume of AI content makes manual or automated verification a monumental task.
Who is affected by this gap
For ordinary internet users, the risk is real. Misinformation — whether political propaganda, fake news images, or manipulated videos — can spread faster than ever. Without reliable detection, people may struggle to trust what they see online.
Journalists, fact-checkers, and social media platforms are also affected. They need robust tools to verify content at scale, but watermarking alone does not provide that. It is a label, not a verification system.
Google’s response and partnership push
Google has been actively expanding SynthID partnerships, integrating it into more of its own products and encouraging third-party platforms to adopt it. The company has positioned the technology as part of a broader responsible AI framework.
However, Google has also acknowledged that SynthID is not a complete solution. The company has called for industry-wide collaboration and regulatory support to address the misinformation challenge holistically.
What the research says about watermark effectiveness
Studies on watermarking technologies show that while they can be effective against casual tampering, determined adversaries can often find ways to bypass them. SynthID’s resilience is a step forward, but no watermark is unbreakable.
Researchers emphasize that watermarking must be part of a multi-layered strategy that includes content provenance, authentication standards, media literacy, and platform enforcement.
Confirmed facts vs what remains unclear
Confirmed: Google’s SynthID is designed to be tamper-resistant and is being expanded through partnerships. The scale of AI-generated content is massive — over 100 billion images and videos from Google alone.
Unclear: How effective SynthID is against sophisticated tampering in real-world scenarios. Whether widespread adoption will happen voluntarily. Whether watermarking alone can meaningfully reduce misinformation at scale.
Why Google’s approach matters beyond the technology
Google’s push for SynthID reflects a broader industry trend: tech companies are investing in technical solutions to problems that are fundamentally social and regulatory. Watermarking is necessary, but it is not sufficient.
The company’s market position — with billions of users across Search, YouTube, and Android — gives it unique leverage to set standards. But that same scale also means any gap in detection can have outsized consequences.
Risks and balanced view
Critics argue that over-reliance on watermarking could create a false sense of security. If users assume all watermarked content is safe and all unmarked content is suspicious, they may be misled by unmarked fakes or trust watermarked content that is still misleading.
There are also concerns about privacy and surveillance. Watermarking systems that track content could be misused for censorship or monitoring.
The wider pattern: Tech’s detection arms race
SynthID is part of a larger trend where AI companies build both the tools that generate synthetic media and the tools that detect it. This creates an inherent tension — the same companies profit from AI adoption while trying to mitigate its harms.
Regulators in the EU, US, and India are beginning to explore mandatory watermarking requirements. But enforcement remains a challenge.
Practical guidance for readers
For now, users should not rely solely on watermarks to judge content authenticity. Cross-check information with trusted sources, use reverse image search, and be skeptical of highly emotional or sensational media.
Platforms should adopt multiple detection methods, including SynthID, but also invest in media literacy campaigns and transparent content policies.
Future outlook
SynthID will likely become more common across Google’s ecosystem and beyond. But without mandatory adoption and complementary detection systems, it will remain a partial solution. The misinformation challenge will require coordinated action from tech companies, governments, educators, and users.
Our Take
Google’s SynthID is a technically impressive tool that addresses one piece of a much larger puzzle. It is hard to break, but it does not solve AI misinformation. The real work lies in building trust systems that combine technology, policy, and human judgment. Until then, the gap between AI’s creative power and our ability to verify reality will continue to widen.
Frequently Asked Questions
What is Google’s SynthID watermark?
SynthID is an invisible digital watermark embedded into AI-generated images and videos. It is designed to survive common edits like cropping and compression, making it harder to remove.
Can SynthID stop AI misinformation?
No. SynthID only labels content after creation. It does not prevent generation of misleading content, nor does it work if the watermark is removed or if a different AI tool is used.
Is SynthID mandatory for all AI content?
No. Adoption is voluntary. Google is expanding partnerships, but many AI tools do not use SynthID, and bad actors can avoid it.
What should I do to avoid AI misinformation?
Do not rely solely on watermarks. Cross-check information with trusted sources, use reverse image search, and be cautious of highly emotional or sensational media.
