AI-watermarked content: what it means for your company's SEO

On October 5, 2026, OpenAI announced it will start watermarking the text that ChatGPT and Codex generate for users in the European Union with a statistical watermark called textGrain — the first real-world rollout of the EU AI Act’s transparency rules by a provider of this size. For any business publishing AI-written content or code — practically everyone, at this point — the immediate question isn’t technical, it’s practical: does this affect my website’s SEO, my marketing content, or the software a developer is building for me? At Evicron, an AI and custom software studio based in Barcelona, here’s what actually changes and what’s worth checking this week.

What OpenAI announced

According to TechCrunch and Unite.AI, OpenAI will turn on watermarking by default, over the coming weeks, for text generated by ChatGPT and Codex for EU accounts, across every plan. The watermark, called textGrain, is invisible: it subtly nudges the model’s word choices against a secret statistical key that a detector can test for, but that a normal reader can’t notice. For API customers outside the EU, watermarking stays off by default and opt-in — each developer decides whether to turn it on.

OpenAI has also opened applications for access to its watermark detector, for now limited to researchers and evaluating organizations, in line with the EU’s Code of Practice on transparency of AI-generated content.

Why: Article 50 of the EU AI Act

This obligation isn’t new, or specific to OpenAI: Article 50 of the EU AI Act, in force since August 2, 2026 (see our earlier piece on the transparency rules that landed in August), requires synthetic content — text, image, audio or video generated by AI — to be identifiable as such through technical means. What OpenAI is doing is the first visible, large-scale implementation of that requirement for text, a format that’s far harder to watermark reliably than an image or a video. Text is, by a wide margin, the hardest of the four formats Article 50 covers to audit: unlike an image carrying C2PA metadata, a statistical watermark in words loses reliability the moment the text gets edited, translated or paraphrased — something several technical analyses of the system already point out. In Spain, it will be AESIA — the agency taking shape to oversee these obligations — that has to deal with that practical limitation.

What it means for your company’s content and SEO

If you publish AI-generated content on your blog or website

OpenAI’s watermark doesn’t, by itself, change how Google or generative AI engines evaluate your content: that still comes down to usefulness, accuracy and genuine authorship, not whether an invisible mark is present. What does change is the regulatory backdrop: if your business uses ChatGPT to write blog posts, product descriptions or SEO content, that text will start carrying a technical mark that third parties can audit. It doesn’t require disclosing anything visible on your site, but it’s worth knowing exactly which parts of your content are generated, edited or fully human-written — that traceability is precisely what Article 50 asks for, and it’s the same point we made about what your website needs for AI to recommend it: verifiable content with real editorial judgment behind it, not just generated and published without review.

If you use Codex or another coding assistant

For code, the practical impact is smaller — nobody expects “watermarks” in software delivered to a client — but the underlying logic is the same: knowing which part of a build is AI-generated and who reviewed it matters more every year in audits, contracts and technical due diligence. It’s one of the reasons custom software projects at Evicron always ship with documented code the client owns outright, with no lock-in to a specific tool.

What to check now

Before watermarking reaches your own ChatGPT account, three things are worth reviewing:

1. How much of your published content comes from AI with no editorial review. It isn’t a legal issue yet, but it’s a quality signal Google penalizes the same way with or without a watermark.
2. Whether your AI provider tells you which content is synthetic. It’s one of the questions we recommend asking when choosing an AI consultancy: who answers if AESIA asks for traceability.
3. Whether your content policy distinguishes between an AI draft and a published piece. A clear editorial standard protects both your SEO and your compliance, and it’s cheaper to put in place now than to fix later.

OpenAI’s watermark is a first step, not the end of the conversation: Anthropic and Google will need to take equivalent steps to operate in the EU, and it’s reasonable to expect the detector to open up beyond its current circle of researchers in the coming months. In the meantime, the best defense is the same one it’s always been: useful, verified content with a clear author behind it.

Want to check whether your company’s content or software is ready for the EU’s transparency rules? Get in touch: the first session is free and we reply within 24 hours.