AI visibility and GEO·September 12, 2026·3 min readLire en français →·By Geneviève Cyr

Claude's AI text watermark: a signal, not proof

In August 2026, Anthropic announced a watermark built into the text Claude produces, applied everywhere in the world. It is not an AI content detector: the company itself writes that the mark is a signal and not proof. And the disclosure obligation that does exist is European, with an exception for content that goes through a human review someone takes responsibility for.

Three facts, and a conclusion that runs against the prevailing worry.

What the mark says and what it does not say

The watermark is an imperceptible signal inserted into the text produced by models launched from August 2, 2026 onward. It is not visible, it changes neither the meaning nor the readability of the answer, and it applies worldwide rather than to European users only.

Anthropic describes the scope of this mark in its own documentation, and the wording is worth repeating as is: it is a signal, not proof. Three limits follow, all written by the company itself.

  • A marked text may have been written by a person: asking the model to proofread or translate a paragraph is enough to leave a trace.
  • A substantial revision erases the signal: a text reworked in depth stops being detectable.
  • An unmarked text is not human for all that: a short passage provides too little material, and other models do not carry this mark.

A tool that claims to detect AI-written content from this signal therefore goes beyond what the signal allows you to conclude, in both directions.

The framework that really matters is elsewhere

Article 50 of the European Artificial Intelligence Act has applied since August 2, 2026. It targets text published for the purpose of informing the public on matters of public interest. A service page, a product page or a commercial blog post does not fall into that category.

The text also provides an explicit exception: the obligation does not apply when the content has gone through human review or editorial control and a person holds editorial responsibility for the publication. An organization that names an editorial lead and documents its review places itself outside the obligation.

Finally, it is a European obligation. It affects a Quebec business only if that business addresses the EU market. No Canadian or Quebec law currently imposes an equivalent disclosure on marketing content.

To decide

What this changes in your practice

  • If you do not target the European market: nothing is required of you legally.
  • If you target the European market: name an editorial lead and keep a record of the review, which is enough for the exception.
  • In every case: keep having each publication signed off by a person who answers for it, because that is what protects quality as much as compliance.
  • What not to do: buy an AI content detection tool to settle an editorial question, since the signal does not allow a conclusion one way or the other.

A vendor who promises to make your content undetectable is selling an answer to a problem that is not yours. What decides your ranking is the quality of what you publish.

Setting who reviews, who signs off and what gets documented is the kind of framework we put in place in a consultation.

Worth noting

Google's position on the subject has not changed with the watermark: what counts is the quality of the content, not the process that produced it. The subject belongs to compliance and editorial governance, not to SEO.

What Google actually does with AI-written content is covered in our analysis on whether Google penalizes AI content. The writing method that holds up is in AI-assisted writing without losing your rankings and the internal framework to put in place is in the AI usage policy for SMBs.

Sources and references
  1. Anthropic, How Claude marks AI-generated content, official documentation, August 2026.
  2. Article 50 of the European Artificial Intelligence Act, applicable since August 2, 2026.
  3. European Commission, transparency obligations under Article 50, accessed September 2026.
Geneviève Cyr
Geneviève CyrPartner · Web development, SEO and GEO

Geneviève puts the strategy for your engagement into action. She leads all our web development projects: Shopify, WordPress and the new ways of building a site with AI. She manages our team of developers and translates your business needs into technical language. She runs your organic search (SEO), your visibility in AI answers (GEO) and your site's conversion rate optimization (CRO). Her work is at the heart of three goals: Attract customers with SEO and AI, Improve your site's conversion, and Strengthen your visibility in AI answers. With Gabriel, she also builds the landing pages for your advertising campaigns. She writes mainly about SEO, AI visibility and web design.

About Falia →

Keep reading

Tout AI visibility and GEO →
01
AI visibility and GEO·10 min read

What AI understands and says about your business, and how to check it

02
AI visibility and GEO·10 min read

Get cited by ChatGPT and other AIs without buying links

03
AI visibility and GEO·3 min read

GEO is 80% SEO: what the remaining 20% changes

Other topicsStrategySEOPaid advertisingConversionAI visibilityWeb design
← All insights