How Claude’s New Watermarking will Affect Indie Publishing
Image courtesy of Claude.com
On August 2, 2026, Anthropic, the makers of Claude, with little to no notice went live in a Claude Help Center article announcing its new “Watermark” feature, which they will be implementing in all models released after August 2. It is rumored OpenAI is expected to follow suit with ChatGPT but a press release has yet to be issued. The new addition is being seen as an unwelcome change that could either usher in the demise of or herald a new dawn for those involved with indie publishing.
The new supplement to their code base is an attempt to comply with the EU AI Act’s Transparency Code that in part states:
Providers must:
Design AI systems in a way that ensures individuals are explicitly informed whenever they interact with an AI system directly
Add machine-readable marks to enable the detection of AI-generated or manipulated content
Deployers of AI systems must inform individuals when they are exposed to:
Emotion recognition and biometric categorisation tools
Deepfakes
Text publications on matters of public interest without human review or editorial control
The Act goes on to state “These obligations are intended to foster trust and integrity in the information ecosystem. People should know when they are interacting with AI or exposed to AI-generated content. This will help them make informed decisions, calibrate their trust and reliance on AI and avoid misinformation or deception.”
What if I just Copy and Paste?
In this specific use case, AI watermarking doesn’t work in the same way image watermarking does. The code model Anthropic is executing is done, not only at the token-level, but also built into the output text mathematically and differs with each instance. It is based on the length of the output and the word choices made by Claude itself. This eliminates the possibility of simply using copy and paste or even re-writing the entire output into a text editor since the text itself will be noted as written by AI.
What does this mean for Indie Authors and Publisher?
Regardless of which side of the spectrum you’re on, this will hit the indie community hard. First, misrepresenting authors will not be able to use AI to write any part of their work without immediately being flagged. On the other side of the sword it also dictates legitimate authors may not use AI for grammar or spell checking as any results produced by the AI will have the same outcome.
Which leads directly into how this will widen the margins inside the indie community. Surprisingly what I believe is not going to be damaging is the level or style of future work being produced since this new course change will correct itself eventually. What will be more affected is the work that has already been published or in the process of being commissioned. Watermarking will ultimately expose those using AI for writing and passing it off as their own by forcing them to turn in work that isn’t produced at the same level either grammatically, through vocabulary, or the train of thought as their AI counterparts.
Flipping the coin we can see this being a boon in the indie publishing realm. Small publishers usually don’t have the resources to verify each submission as they come in. And, while there are software suites that can detect AI writing, most of them are either semi-accurate at best or financially out of reach. This is how many non-writers are able to get their work past smaller houses with little to no acquisition editors.
Having the ability to discern right out of the box whether a piece was written genuinely by a human author or not, can be a cost and time saving measure for most indie houses and magazines. It will eliminate the need for the necessary yet often costly detection software and free up more time for other submissions that don’t need as much background verification.
What should Authors do Moving Forward?
None of this should frighten authors from using AI, but instead help them move to better ways of using it. For now, there seems to be two ways for authors and publishers to define the future of AI usage:
Use AI for research.
AI is already adept at finding details that would take traditional research methods weeks to months to complete. Utilize this speed of action to help find the information you need to design your world building for instance, “what items were sold at a normal French market in 1773?”.Learn to Paraphrase.
In both undergrad and grad school, having the ability to paraphrase—a restatement of a text, passage, or work giving the meaning in another form—according to Miriam Webster, was a super power that helped me understand fiction and nonfiction in a way that I could then explain it to anyone later. You have to understand, fundamentally, anything written by AI is not your work. It is based on an algorithm that has taken the works of millions of others and presented to you based on a proprietary predictive model. As a writer you need to learn to rephrase anything that AI outputs into your own words and combine it with your own ideas. This has been the basic foundation of the craft for generations, and as the old saying says... “if it ain’t broke....”
When it’s all said and done, AI is simply another tool. It is not meant to replace honest hard work and true human writing, but to supplement it. While there are talks of the AI bubble bursting, no one is sure of the exact date and time of that happening. So, if you’re in the indie publishing sphere, it’s best to learn how AI works and to be able to detect its markers. Because, in the end, true and authentic human writing will always win out over the uncanny valley that is AI generated text.
