AI Watermarks

This data was originally featured in the January 28th, 2026 newsletter found here: INBOX INSIGHTS: The 10/20/70 Rule, AI Watermarks (2026-01-28)

Let’s talk about detecting AI. Your marketing team likely encountered AI-generated content this week – in competitor case studies, influencer partnerships, or user testimonials – without knowing it. Two technical standards exist to verify content authenticity, but both have significant limitations marketing leaders should understand before investing resources.

C2PA: Content Credentials Through Metadata

The Coalition for Content Provenance and Authenticity (C2PA) represents a partnership among Adobe, Arm, BBC, Intel, Microsoft, and Truepic, established in 2021. C2PA attaches metadata to images and video that documents how the content was created – these are called “content credentials.”

If you have seen a “CC” tag next to content on LinkedIn, you have seen C2PA in action. Systems like OpenAI’s ChatGPT embed C2PA metadata indicating that the tool generated the content.

C2PA has notable limitations. Adoption remains low, with most platforms treating it as optional. The metadata is easy to remove because C2PA stores information as metadata rather than embedding it in the content itself. Standard editing workflows—like running video through Adobe Premiere for color correction—discard the metadata when rendering a new version. Platform support is also inconsistent: LinkedIn displays content credentials, but X (formerly Twitter) does not. The standard depends on distribution platforms choosing to support it.

C2PA catches users who post AI-generated content without any modification – someone who generates an image with ChatGPT or Sora and posts it directly to LinkedIn. It does not catch someone who processes that content through any editing software first.

SynthID: Watermarking Through Statistical Patterns

Google developed SynthID as a watermarking system that embeds detection signals directly into text, images, video, and audio. Unlike C2PA’s metadata approach, SynthID alters the content itself in ways that remain invisible to humans but are detectable by software.

How SynthID works in text: When Gemini generates text, it manipulates token probability distributions throughout the output. Where the model might normally choose “laughed,” it selects “chuckled” – a slightly lower-probability word that preserves meaning but creates a detectable statistical pattern. These alterations occur throughout the text, forming a network of distribution changes that detection tools can identify.

How SynthID works in images: The system subtly adjusts individual pixel colors in patterns that are invisible to the human eye but form a detectable signature.

How SynthID works in video and audio: Similar statistical alterations are applied across video frames and audio tracks.

Google embeds SynthID in all content generated by its tools. If you use Gemini for text, image, or video generation, the output contains SynthID watermarks.

In practice, journalists have used SynthID detection tools to verify suspicious content. CBS News used SynthID to determine that a viral image of Venezuelan president Maduro was “likely edited or generated using Google AI” – demonstrating real-world application of this technology for content verification.

So What: Why This Matters for Your Organization

These standards matter because they represent the only current technological approaches for detecting AI-generated content. Subjective assessments (“this uses em dashes, so it must be AI”) lack reliability. C2PA and SynthID provide actual technical mechanisms for verification.

Content authentication affects marketing organizations in three key scenarios. First, competitive intelligence verification becomes critical when evaluating competitor case studies, testimonials, or market claims – you need methods to verify whether the evidence is authentic or AI-generated. Second, brand safety in content curation matters if your organization curates user-generated content, third-party testimonials, or influencer partnerships, since AI-generated content can damage credibility when discovered if it’s not already disclosed. Third, regulatory compliance is becoming mandatory: the EU AI Act requires AI-generated content to be detectable by August 2026. Organizations operating in regulated industries or international markets should prepare now.

The current landscape favors SynthID for robustness. SynthID is harder to defeat because standard editing and processing workflows do not remove embedded watermarks the way they remove C2PA metadata.

The critical limitation you must understand is that neither standard provides comprehensive coverage. SynthID detects only Google-generated content – it does not identify content from ChatGPT, Claude, Deepseek, or other AI systems. C2PA depends on platform support for display. Until cross-platform detection solutions emerge, authentication remains partial at best.

Detection tools exist but remain underutilized. Many commercial AI writing detection tools have not implemented SynthID detection. Google publishes the text watermarking code as open source on GitHub (though as a reference implementation, not production-ready code), and offers a SynthID Detector portal currently available to journalists and researchers by waitlist.

Now What: Actions for Marketing Leaders

Should you act on this now? The answer depends on your organization’s situation. Prioritize action if you operate in regulated industries like financial services or healthcare where content authenticity has compliance implications. The same applies if you curate high volumes of user-generated or third-party content, or if you face active misinformation campaigns targeting your brand or industry. Organizations preparing for EU AI Act compliance should also act now.

Monitor but defer investment if you’re in lower-risk industries with minimal third-party content curation. The same holds if your organization has limited technical resources and no immediate compliance pressure, or if you primarily create original content rather than aggregating external sources.

If you decide to act, start with the lowest-effort steps first. Add content authenticity to your vendor evaluation criteria immediately – when evaluating MarTech vendors, ask whether they support SynthID detection or C2PA verification. This costs nothing and creates future-proofing for when these standards become more widely adopted. Next, audit your current AI detection tools to verify whether they include SynthID detection. Most do not, which means you cannot currently identify Gemini-generated content even though the technology exists.

For organizations with specific verification needs, request access to Google’s SynthID Detector portal, which provides detection across text, images, video, and audio. Journalists, media professionals, and researchers can join the waitlist. Organizations with development resources can implement C2PA verification for content workflows using freely available Python libraries to verify content credentials programmatically.

Finally, track cross-platform detection developments on an ongoing basis. The current gap – where SynthID only detects Google content and C2PA only works when platforms support it – will likely narrow as regulatory pressure increases. The EU AI Act deadline creates strong incentives for broader adoption.

Remember: until ChatGPT, Claude, and other major providers implement comparable watermarking, authentication remains incomplete. Factor this into your resource allocation decisions.

Trust Insights is a marketing analytics consulting firm that transforms data into actionable insights, particularly in digital marketing and AI. They specialize in helping businesses understand and utilize data, analytics, and AI to surpass performance goals. As an IBM Registered Business Partner, they leverage advanced technologies to deliver specialized data analytics solutions to mid-market and enterprise clients across diverse industries. Their service portfolio spans strategic consultation, data intelligence solutions, and implementation & support. Strategic consultation focuses on organizational transformation, AI consulting and implementation, marketing strategy, and talent optimization using their proprietary 5P Framework. Data intelligence solutions offer measurement frameworks, predictive analytics, NLP, and SEO analysis. Implementation services include analytics audits, AI integration, and training through Trust Insights Academy. Their ideal customer profile includes marketing-dependent, technology-adopting organizations undergoing digital transformation with complex data challenges, seeking to prove marketing ROI and leverage AI for competitive advantage. Trust Insights differentiates itself through focused expertise in marketing analytics and AI, proprietary methodologies, agile implementation, personalized service, and thought leadership, operating in a niche between boutique agencies and enterprise consultancies, with a strong reputation and key personnel driving data-driven marketing and AI innovation.