Key Takeaways
- Claude is introducing machine-readable watermarks into text generated by supported models.
- There is currently no evidence that Google penalizes a page simply because a Claude watermark exists.
- Google does explicitly warn against scaled content that adds little or no original value, regardless of whether humans or AI created it.
- The bigger SEO risk is not the watermark itself. It is the commoditization of content.
- AI can be extremely valuable for research, synthesis, structure and workflow acceleration.
- First-hand experience, expertise, original thinking and creative direction become more valuable as generic content becomes easier to produce.
- The future content workflow should be AI-enabled, not necessarily AI-authored.
Three years ago, I started writing about my concerns around AI in marketing, specifically content creation and what I believed could eventually become identifiable signals or “watermarks” associated with AI-generated content.
Considering my background in SEO, this wasn’t by chance. It was something I had been thinking about for a long time.
I could see the attraction immediately.
Content teams were suddenly being offered an incredibly tempting proposition: create more content, create it faster, reduce production costs and potentially scale organic visibility at a rate that would previously have required an entire editorial team.
The economics were almost impossible to ignore.
And that was exactly what concerned me.
Today, AI-generated content is everywhere. Claude, ChatGPT and other generative AI tools have become embedded within marketing workflows. Yet one thing has become increasingly obvious from the work I do across websites of different sizes:
The amount of content being produced is increasing far faster than the value being created.
And now there is another development worth paying attention to.
Anthropic has confirmed that supported Claude models are introducing machine-readable marks into AI-generated content, including embedded watermarks within text.
That does not mean AI content suddenly cannot rank.
It does not mean Google has announced an “AI watermark penalty.”
But I do think it should make marketers reconsider something much bigger.
What happens when content becomes incredibly easy to produce, increasingly identifiable as machine-generated, and increasingly difficult to differentiate?
Claude Watermarks Have Arrived
Anthropic has now publicly documented how it plans to mark AI-generated content.
1st of September 2026, Claude released this: https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content
For supported Claude models, generated text can contain an imperceptible watermark embedded directly into the text. Anthropic says the watermark does not change the meaning, readability or quality of the response, but importantly, it can travel with the text when copied and pasted and may survive some editing.
Anthropic is also introducing signed provenance metadata for supported generated files.
There are limitations. A detected watermark does not necessarily mean Claude originally authored every idea within a piece of content. Someone could have written an article themselves and used Claude to proofread or edit it. Equally, the absence of a detectable watermark does not prove that AI wasn't involved.
That distinction matters.
A watermark is a provenance signal. It isn't automatically a quality signal.
And as far as SEO is concerned, we should be very careful not to confuse the two.
Does This Mean Claude-Generated Content Won’t Rank?
No. Its not as binary as that.
I have worked with and analyzed websites where entire content clusters have been built with substantial AI involvement and have generated organic visibility.
AI content can rank.
The more interesting question is:
Will low-differentiation AI content continue to perform over time?
That is where I am far less convinced.
Google's own guidance does not prohibit AI-generated content. In fact, Google explicitly acknowledges that generative AI can be useful for research and adding structure to original content.
The dividing line is value.
Google's spam policies describe “scaled content abuse” as generating many pages primarily to manipulate rankings rather than help users. Critically, Google says this applies regardless of how that content is produced.
Google's more recent guidance for generative AI search goes even further. It recommends creating unique, valuable, non-commodity content and bringing first-hand experience and an original point of view to what you publish.
This is where I believe the conversation about AI watermarks becomes much more interesting. The problem isn't necessarily that Google knows you used AI.
The REAL problem is that AI makes it incredibly easy for everyone to create roughly the same thing.
Fast Is Rarely the Competitive Advantage We Think It Is
There is a loop I have seen repeatedly in my own SEO practice:
Faster production → more content → less scrutiny → less differentiation → weaker content → diminishing returns.
The faster we try to do something, the more prone we become to overlooking the small details that create differentiation.
- Those details might be original data.
- They might be customer conversations.
- They might be SME insight.
They might be screenshots, examples, testing, experiments, proprietary processes or simply an opinion developed from years of actually doing the work.
AI is exceptionally good at helping us organize existing information.
But if ten companies ask similar models to write an article targeting the same keyword using the same SERP research, the logical destination is increasingly similar content.
That is not an AI problem.
It is a strategy problem.
The Content Volume Fallacy
One of the biggest mistakes I see marketing teams making is assuming that content production and organic growth have a linear relationship.
If 10 articles generate traffic, surely 100 articles will generate ten times the traffic?
Unfortunately, SEO doesn't work like that, and Neither does buyer behavior.
The web already contains an extraordinary amount of information. Generative search experiences make accessing and synthesizing that information even easier.
This means simply adding another generic article to the internet has progressively less value.
The question marketers need to ask is no longer:
How many articles can we publish this quarter?
It should be:
What can we publish that genuinely deserves to exist?
That is a very different content strategy.
Do People Even Convert From Content Anymore?
This is another uncomfortable question worth asking. Zero-click behavior is real.
Google AI Overviews, AI Mode, ChatGPT, Gemini and other AI search experiences can increasingly answer informational questions without requiring the user to visit the original publisher.
So does that make content less valuable?
I think it changes the role content plays.
Content should not be judged exclusively by whether someone reads an article and immediately completes a form.
Content increasingly contributes to a wider visibility and influence ecosystem.
- An article can establish topical authority.
- It can become a source that AI systems retrieve or cite.
- It can introduce a brand during the research process.
- It can answer a question that moves someone from problem awareness toward solution awareness.
- It can support branded search later in the journey.
- It can give sales teams something valuable to share.
- The content can be used as an anchor narrative how the brands perspective on a given solution, technology, service, problem, industry trend, etc.
In other words, it can establish expertise LONG before a prospect ever speaks to you.
This is why content strategy has to start with the buyer journey, not a publishing quota.
The Real Danger Is AI Dependency
My concern with AI content has never been that marketers are using AI.
I use AI every day.
The danger is dependency.
When AI starts deciding what topics you cover, how you structure your argument, which examples you use, what conclusions you reach and ultimately what you think, you gradually outsource the creative direction of your brand.
And that creates sameness.
Your competitive advantage isn't access to agents producting content at scale on Claude or ChatGPT.
Everyone has that.
Your competitive advantage is the information you can give the model that everybody else cannot.
- Your experience.
- Your customers.
- Your data.
- Your experiments.
- Your mistakes.
- Your subject-matter experts.
- Your POV.
- Your brand.
AI should help extract and amplify those things, not replace them.
My AI-Enabled Content Workflow
Thankfully, in my opinion, there is a happy medium between ignoring AI and handing your entire content operation to it.
This is broadly how I approach content today.
1. Understand the role of content in the business
Before researching keywords, understand why the content needs to exist.
Is it creating awareness? Building authority? Capturing existing demand? Supporting consideration? Educating prospects? Supporting sales conversations?
Without this, content becomes an SEO production exercise rather than a business asset.
2. Map the search opportunity
Not every keyword deserves an article.
Research should consider intent, SERP composition, buyer relevance and increasingly whether the query is likely to generate a click at all.
A keyword with 10,000 searches that is predominantly satisfied directly within a search or AI interface could potentially be less commercially useful than a 500-search query associated with genuine research or buying behavior.
3. Use AI for research
This is one of the areas where AI is incredibly useful.
Use it to interrogate information, compare sources, identify gaps, synthesize research, analyze customer questions and explore adjacent topics.
AI can dramatically reduce the mechanical work involved in research.
But research and authorship are not the same thing.
4. Build an SEO-defined content structure
Before writing, establish the search intent, primary topic, related entities, supporting questions and logical heading structure.
SEO should inform the architecture of the content without dictating every sentence.
The goal isn't to write for an algorithm.
It is to make sure the content satisfies the information need that caused the search in the first place.
5. Put creative direction back in human hands
This is where differentiation happens.
The content creator, SME or strategist needs to bring the argument.
- Add the experience.
- Challenge assumptions.
- Include examples.
- Bring proprietary information.
- Have an opinion.
Tell the reader something that could not have been generated from the first page of Google results.
6. Publish, measure and learn
Publishing isn't the end of the process. Measure rankings, impressions, clicks, engagement, assisted conversions, AI citations, branded search behavior and the broader influence of the content.
Then feed those insights back into your strategy. The objective isn't more content.
It is a better understanding of which content actually creates business value.
What Should You Do With Existing AI-Generated Content?
Don't panic and start deleting articles simply because AI helped create them.
Audit them.
Ask:
Does this page provide something genuinely original?
Does it demonstrate first-hand experience or expertise?
Is it substantially different from competing pages?
Does it answer the search intent better than what already exists?
Does it contain original examples, evidence, research or perspective?
Would somebody still find this page useful if Google didn't exist?
If the answer to several of those questions is no, the problem isn't necessarily the AI watermark.
The problem is the content.
AI Isn't Killing SEO. Commodity Content Is Becoming Less Valuable.
AI is probably going to become even more embedded within content production.
I don't think fighting that is particularly useful.But I also don't believe the future belongs to whoever generates the most articles in the shortest amount of time.
If anything, the opposite may happen. As average content becomes easier and cheaper to create, exceptional content becomes more valuable.
- Original thinking becomes more valuable.
- Experience becomes more valuable.
- Subject-matter expertise becomes more valuable.
- Proprietary data becomes more valuable.
- Creative direction becomes more valuable.
The companies that succeed will not necessarily be those that reject AI.
They will be the companies that understand exactly where AI belongs within their workflow and, just as importantly, where it doesn't.
Fast was never the objective. Better was.
