You’ve seen the headlines. Your AI content is watermarked now. And if your first thought was that you used ChatGPT on your About page back in March and now you’re in trouble, this episode is for you.
We break down what an AI watermark is, whether AI-generated content hurts your SEO, what Google has said on the record since 2023, and whether you can (and should) get AI watermarks removed.
Short version: A watermark tells you a tool was in the room. It does not tell you whose idea it was, and it is not what decides your rankings.
What we cover in this episode:
- How we use AI in our own content, and exactly where we stop
- What an AI watermark is, and the three things people keep mixing up
- Why AI detectors and AI watermarks are two completely different animals
- Why ChatGPT’s new watermark only applies in Europe (for now), and what that means for you
- Whether Google penalizes AI content, in Google’s own words
- Who owns AI-generated content when a machine helped write it
- What the experts at NP Digital said on camera about rankings
- The one claim going around that the source documentation contradicts
- Three things to do this week
How We Use AI at Duo Collective
We want to be honest, and we’ve never been shy about saying we use AI in our business.
Here’s exactly how we use it:
- Brainstorming content ideas
- Outlining our content structure (think headlines)
- Verifying that the voice and tone match
- Proofreading
- Integration with Semrush to pull and integrate keywords
- Finding missing holes or spots where our content can do better
- Finding missing GEO opportunities
We have not one, but two human copywriters on our team for a reason.
What The Heck is an AI Watermark?
An AI watermark is an invisible signal woven into text as a model generates it, so a system holding the right key can later check whether that model was likely involved.
Nothing gets stamped on the page, and there are no hidden characters.
Anthropic (the creator of Claude) said, “The difference between watermarked and unwatermarked text is not distinguishable to readers, and the mark carries no information that identifies a person, an organization, or a chat. It exists because the EU’s AI Act requires providers serving that market to mark AI-generated content, and around 190 organizations signed the same code of practice, so this is a compliance move rather than a punishment.”
How does the watermark work?
The AI builds a sentence one word at a time. A lot of the time, two words would work just as well. Overcast or grey. Stocky or sturdy. Nobody reading it would notice either way.
So the AI has to pick one. It used to pick at random, like flipping a coin.
Now it uses a secret cheat sheet instead. The cheat sheet says which of the two to grab. You can’t see the cheat sheet. Your reader can’t either.
But the company that made the cheat sheet can go back later, read a long chunk of text, and check how many of those little choices match what their cheat sheet would have said. One match means nothing, because a person could pick “grey” by accident. Hundreds of matches in a row mean the AI wrote it.
That’s the whole watermark. It is not a stamp on your work. It is a pattern hiding in which words got picked.
Which leads to the part that matters for you. The cheat sheet only gets used when the AI is the one choosing a word. If you wrote the thing yourself and handed it over for a grammar fix, the AI barely picks anything. Your words are already sitting there. There’s almost nothing for the pattern to live in.
AI Detectors are the Real Problem
The watermark isn’t the thing that’s going to hurt you. It’s actually going to be running your copy through an AI checker (ZeroGPT, GPTZero, QuillBot, Grammarly, etc.), which is a completely different technology from AI systems (Claude, ChatGPT, Gemini, etc.).
How AI Detectors Work
An AI detector never sees how your content was made. It reads the finished words from the outside and guesses.
What it’s measuring is called perplexity, which is a fancy word for how predictable your word choices are. Predictable writing gets scored as AI. So it’s not measuring whether a machine wrote it. It’s measuring whether it reads the way machines read.
A real watermark detection tool exists. Anthropic has it in private preview, and it’s only available to regulators, law enforcement, media, fact-checkers, researchers, educational organizations, EU civil society groups, and companies with their own legal obligations.
So if somebody accuses you of writing content with AI, they’re holding the guessing tool. A watermark is evidence. A detector is an opinion.
Is Your Content Marked?
The watermark only applies to the words the AI chooses. You write it, hand it over, and ask for a grammar clean-up, but nearly every word is still yours. There may be almost nothing for the mark to grab onto.
Even if your content is watermarked… here are three things it does NOT do:
- Carries nothing that identifies you, your business, or your chats
- Does not change who owns it or who’s responsible for it
- Cannot tell anybody how much you did versus how much it did
ChatGPT said the same three things Anthropic did. When two of the biggest AI companies are saying the exact same thing, you can relax a little.
Can I get Watermarks Removed?
People are already Googling how to remove these. There are twelve thousand searches a month for watermark removers.
Butttt… don’t bother.
IEEE Spectrum said: “Stripping the mark out of a long piece takes changing about a quarter of the words or more.”
You don’t need a tool. You need to edit like you mean it. If you are shopping around for a remover, the watermark isn’t your problem… crappy content is (sorryyyyyy, there’s no shortcut here.)
ChatGPT tested their own watermark, and when they swapped out about a quarter of the words, it mostly disappeared.
Heavy editing does the job, and you don’t need a sketchy tool for that.
Do I Have to Disclose That I Used AI?
Short answer for almost everyone listening: no.
EU AI Act Article 50 requires disclosure for AI text published to inform the public on matters of public interest. That’s news reporting, not your blog or your service page. And even there, the requirement falls away if a human reviewed it and a named person holds editorial responsibility.
Google has no disclosure requirement and no ranking penalty for staying quiet about it.
Google’s guidance does say to use AI disclosure when readers would reasonably expect to know how the content was produced.
We’re not lawyers, so check what applies where you operate.
Nobody is coming after you for your blog post. The person who can come after you is the client paying you to write. That’s a contract problem, not a search problem.
Do This: A Policy Instead of a Label
A label on every post says a machine touched this, and doesn’t tell the reader anything useful. They might just assume AI wrote everything on the page.
A policy says what you use it for, where you stop, and who’s responsible. That’s what people actually want to know.
Here’s what your policy should include:
- What you use AI for, plainly. The same list we gave you at the top of this episode
- What never touches AI. Name the things
- Who holds responsibility. A human reads every word and stands behind it
- Put it on your site as its own page, link it in the footer with your other T&C’s
- Same language goes in your contracts if you write for clients
Here’s What The Experts Are Saying:
Google’s been saying the same thing since 2023, and nobody listens. They focus on “rewarding high-quality content, however it is produced.”
They specifically say: “Appropriate use of AI or automation is not against our guidelines.”
Google also says using automation to crank out content whose main job is to game rankings is a spam violation. This has nothing to do with watermarks.
Is AI Content Bad For SEO? Quick Answers
No. Google’s guidance says it focuses on rewarding high-quality content, regardless of how it is produced, and its FAQ states that appropriate use of AI or automation is not against its guidelines. What gets penalized is using automation to mass-produce content whose main job is gaming rankings.
No. Asked directly in a September 2026 webinar, NP Digital’s Director of SEO Lindsay Shrant said rankings do not drop because of a watermark, and that Google is not penalizing AI; it is penalizing sameness.
The watermark doesn’t change who owns your work, and it doesn’t identify you, your business, or your chats. Copyright is a separate question. In the US, content created entirely by AI with no meaningful human involvement can’t be copyrighted, and the Supreme Court declined to revisit that rule in March 2026. The more of your own ideas, stories, and edits are in a piece, the stronger your claim to it. We’re not lawyers, so talk to one if this matters for your business.
Google can check for its own watermark in text written by its Gemini models. It can’t read the watermarks Anthropic or OpenAI use, because each company keeps its own secret key. And there’s no evidence Google uses watermark detection to rank pages. Even a perfect watermark only shows a tool was involved somewhere, so it can’t tell a fully AI-written post from one where AI fixed the commas.
Skip it. Research covered by IEEE Spectrum found that stripping the mark from a long piece takes changing roughly a quarter of the words or more, which is what real editing does anyway.
So What Should You Do About All of This?
At the end of the day:
- Keep the idea yours. Your client story, your opinion, your numbers, your photos. Make sure your ideas, perspectives, and values are driving the writing from the beginning.
- Make your words survive the draft. Prompt + copy/paste is the problem, and it was the problem before watermarks existed.
- Give people something a prompt can’t make. Client stories, hard data, proof, emotion, personality, etc.
Sameness is the problem, and sameness is the one thing no tool can fix for you, which makes it the only place you get to win.
Your work is still yours. Even if it’s watermarked, the human part is the reason a piece works (always been true).
If sameness is what’s holding your site back, that’s the entire point of SEO on Tap. We’d love to teach you even more about AI and SEO inside the course.
Sources:
- How Claude’s text watermark works, Anthropic, August 2026: https://www.anthropic.com/news/claude-text-watermark
- Google Search’s guidance about AI-generated content, Google Search Central: https://developers.google.com/search/blog/2023/02/google-search-and-ai-content
- AI Models Are Watermarking Text, IEEE Spectrum, September 2026: https://spectrum.ieee.org/ai-watermark-text-anthropic-openai
- The Future of Content Marketing: How AI Watermarking Changes the Game, Neil Patel and NP Digital, September 2026: https://www.youtube.com/watch?v=5nibaAAloZ4
- What is AI watermarking and how should we respond to it, Julian Koplin, Monash Lens: https://lens.monash.edu/its-not-cheating-its-quiet-efficiency-what-is-ai-watermarking-and-how-should-we-respond-to-it/
- OpenAI will start watermarking ChatGPT’s text in the EU, TechCrunch, October 2026: https://techcrunch.com/2026/10/05/openai-will-start-watermarking-chatgpts-text-in-the-eu/
- Supreme Court denies AI authorship claim, National Constitution Center, March 2026: https://constitutioncenter.org/blog/supreme-court-denies-artificial-intelligence-authorship-claim-for-artwork-copyright
- Study of Turnitin and Originality detectors on human, AI, and hybrid texts, International Journal for Educational Integrity, 2026: https://link.springer.com/article/10.1007/s40979-026-00213-1
Listen to the Episode!
Apple Podcasts | Spotify
If you loved today’s episode of The Duo On Air Podcast, we’d love for you to leave a review on your favorite platform and hit subscribe for those new episodes dropping every Monday — it helps more small business owners find us (and let’s be honest, that’s good SEO ).
More From The Duo
Sign Up for Tuesday Confetti Drop Newsletter
Abbey Oslin and Courtney Petersen are Minnesota-based marketing experts with over 20+ years of combined experience working with big brands. They are educators and co-founders of Duo Collective; a boutique organic marketing agency specializing in SEO, branding, and custom websites for women-led small businesses and creative entrepreneurs.
Ready to grow your visibility and become unforgettable?
Learn more or work with our team here.
To inquire about being a guest on Duo On Air, please fill out this application form.
+ show Comments
- Hide Comments
add a comment