The European Union now requires AI companies to watermark AI-generated text. Because these companies operate globally, those compliance mechanisms are being applied worldwide. Claude publicly announced this change in August of 2026, and major AI providers like Google and OpenAI are following suit.
Unfortunately, there’s a lot of misinformation online about how and why AI-generated text is watermarked. That misinformation has caused real concern for independent consultants whose reputations depend on their experience-based expertise.
Will an invisible digital stamp accuse you of passing off AI slop as original thinking?
It might.
But if you adopt an AI Usage Policy disclosing how and when you use AI, and you use generative AI as a tool to refine your thinking rather than replace your independent judgment, technical watermarking becomes a non-issue.
Separating online noise from regulatory reality.
In the United States, none of the federal legislation being debated mandates that AI systems watermark text outputs. Neither does the California AI Transparency Act (SB 942), which applies strictly to image, video, and audio content. Instead, the three acts currently being debated in Congress focus on combatting misinformation:
- The AI Labeling Act requires developers to embed machine-readable metadata into AI-generated images, audio, and video and requires clear disclosures when users interact with AI chatbots.
- The REAL Act requires federal officials to disclose their use of AI when publishing content but exempts text reviewed by agency staff before publication.
- The COPIED Act requires the National Institute of Standards and Technology to develop voluntary provenance standards and protects creators from having their content used to train AI models without consent.
While the U.S. regulations target synthetic media and misinformation, the European Union takes a broader approach centered on transparency.
Article 50 of the EU AI Act went into effect on August 2, 2026. It requires developers to mark AI-generated text outputs with machine-readable metadata that is detectable as artificially generated. Because global AI systems like Claude, ChatGPT, and Gemini operate across jurisdictions, they apply these compliance mechanisms worldwide.
Understanding how text watermarking works.
AI text watermarking is typically done through a process called token selection. Essentially, AI models adjust the mathematical probability of their word choices. That leaves a statistical pattern across the AI-generated text — but that pattern can only be detected if there is a sufficient volume of unedited output.
If you’re using AI to refine your thinking and writing, rather than outsourcing your thinking and writing to the tool, your original copy and revisions to any AI-generated copy will likely break that statistical pattern and render the watermark undetectable. Moreover, detection tools are far from foolproof. They simply indicate that the text may have been generated or manipulated by an AI system.
Turning transparency into a business development asset.
Regardless of whether a tool detects your use of AI, you should disclose it — especially when doing work for a client. At minimum, you should have an AI Usage Policy that clearly states when and how you use AI. (You can review my AI Usage Policy and adapt it for your website and client proposals.)
In addition to an AI Usage Policy, I recommend including an AI Disclosure Statement whenever warranted. Disclosing your use of AI takes all the joy away from those who love to accuse others of using these tools (while referring to them as dinosaurs if they don’t). More importantly, your policy and disclosure statement are valuable business development assets because they reinforce that your judgment and experience-based expertise, not your ability to write a prompt, are what make you an irreplaceable thought partner.
| AI Disclosure Statement: I used NotebookLM to understand how the laws currently in effect and currently being debated impact AI-generated text. I also used a custom AI fact-checking tool I developed in Gemini to fact-check this article. Finally, I used LanguageTool, an AI-based grammar checker, to review this article. To learn more about how (and when) I use AI, review my AI Usage Policy. |
* * *
Erica Holthausen is an authority development strategist and the founder of Catchline Communications. A veteran writer and editor, she equips consultants with the tools to share their perspective through original, high-quality articles that build authority and serve as powerful business development assets. In addition to writing for their blogs, her clients have been featured in Inc., TD Magazine, and Harvard Business Review. To learn how to turn your expertise into recognized authority, join The Authority Lab, a free monthly Q+A for consultants. Register here.
