If your team in Poland writes with Claude or Claude Code, two things changed on August 2, 2026: Anthropic began marking text from its newest models, and Article 50 of the EU AI Act became enforceable. The question behind both is the same: do the marks change how your content ranks on Google.pl, whether it can be promoted at scale, and how LinkedIn and Facebook treat it? Below are the dates, the official statements, the studies with their weak points, and a small baseline from this site, where Claude drafted the pages and I edited, strengthened and curated them. It was drafted with Claude and edited by me.
Short answer
- Will marked content rank lower on Google.pl? No source shows it. Google has published nothing on text watermarks, and no study tested Google.pl specifically.
- Can a Polish team promote AI content at scale? The documented risk is scale, not the mark: Google's spam policies and LinkedIn's stated effort against low quality, automated or generic content apply either way.
- Do LinkedIn and Facebook react to the mark? I found no platform that says it reads Claude's text watermark. The labels they document concern images and video.
- What is different in Poland? The AI Act applies directly. Ask a lawyer how Article 50 applies to your channel.
- What about photos and video? They carry more visible labels: C2PA credentials, SynthID and platform labels on LinkedIn, Meta, TikTok and YouTube. Text is the exception, because no platform is known to read a text mark.
- Our own result: four blog pages that Claude drafted and I edited collected 43 Google organic views and 6 referrals from AI assistants over 90 days, against 402 and 19 for the whole site. Details are in the section "My experience" below.
My experience: four pages Claude drafted, I edited, and their results
I set the topics, edited and strengthened the text, and approved the pages that Claude drafted. In GA4 from June 27 to September 24, 2026, four Google-related pages collected, across all language versions:
| Page | Google organic (views) | AI assistant referrals (views) |
|---|---|---|
| AI Overviews CTR drop | 16 | 0 |
| Search Console, social and video platforms | 11 | 0 |
| Search Console data guide | 10 | 1 |
| Google Ads terms update | 6 | 5 |
The whole site had 402 Google organic views and 19 views from AI assistants in those 90 days. These are small numbers, and all four pages were published before August 2, so they say nothing about the watermark. They are a baseline that I will compare with later pages.
Why Poland is directly affected by Article 50
The AI Act is an EU regulation, so its transparency rules apply to providers and users in Poland without a national law being passed first. Law firm summaries of the Commission's final guidelines list these dates:
- August 2, 2026: Article 50 transparency obligations become enforceable.
- December 2, 2026: grace period ends for machine-readable marking on systems already on the market.
- February 2, 2027: providers must offer watermark detection tools and interoperability.
Article 50(4) also concerns you as a publisher. Deployers must disclose certain AI-generated text on matters of public interest, unless a person exercises genuine editorial responsibility over it, and a routine review is not enough. Whether your blog counts as "public interest" is a legal question, so take it to a lawyer rather than to an SEO article, including this one. Source: Faegre Drinker (https://www.faegredrinker.com/en/insights/publications/2026/7/eu-ai-act-commission-confirms-transparency-code-of-practice-as-adequate-and-publishes-final-version-of-its-guidelines-on-transparency-obligations).
What Claude and Claude Code actually mark
According to Anthropic's Help Center (https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content), text from all supported products carries an imperceptible mark. Claude Code is on that list. Models launched on or after August 2, 2026 mark from launch, and all older models will be covered by December 2, 2026. The marking applies worldwide, so a Polish team gets the same behavior as a team anywhere else.
Some limits worth knowing:
- The mark travels with copy and paste and "may persist through some editing."
- Heavy editing, paraphrasing or translation can remove it. Translation is relevant if you produce Polish and English versions from one draft.
- A detected mark is not conclusive proof of authorship, and a missing mark proves nothing.
- Text detection is in private preview for organizations such as regulators and media, and there is no public checker for text.
Does Google.pl see it? What is known
Google has published nothing that names Claude's watermark, or any text watermark, as a ranking signal. Its stated approach is to reward helpful content and to treat scaled, low-value pages as spam, "regardless of how the pages were made" (spam policies: https://developers.google.com/search/docs/essentials/spam-policies). The SynthID and C2PA verification that Google added to Search in 2026 concerns images.
Three sources, three different readings:
- Ahrefs found that 86.5% of top-ranking pages in a 600,000-page sample had some AI content, with a correlation of 0.011 with position. It is older than the marks (source: https://ahrefs.com/blog/ai-generated-content-does-not-hurt-your-google-rankings).
- First Page Sage reported average position 6 for content without AI and 11 for AI content, but the authors assumed the watermark and did not control quality (source: https://firstpagesage.com/seo-blog/impact-of-ai-watermarking-on-seo-geo-results/).
- Search Institute found no public ranking document that links the mark to positions (source: https://searchinstitute.co/research/does-claudes-ai-watermark-matter-for-seo).
None of them tested Google.pl specifically, so a claim about Polish results has no direct evidence behind it.
AI marks in text, photos and video under EU rules
Article 50 of the EU AI Act is about marking AI-generated content in general, not only text. In practice the technologies differ by format:
- Text gets an invisible watermark (Claude: Anthropic Help Center https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content). No platform in my research says it reads it.
- Photos and video carry C2PA credentials and invisible watermarks such as SynthID. Meta reads C2PA and IPTC on images (https://about.fb.com/news/2024/02/labeling-ai-generated-images-on-facebook-instagram-and-threads/), LinkedIn shows a CR label on images and video (https://fortune.com/2024/11/26/linkedin-content-credentials-labeling-challenges-c2pa), TikTok auto-labels both (https://newsroom.tiktok.com/en-us/partnering-with-our-industry-to-advance-ai-transparency-and-literacy), and Google verifies SynthID and C2PA in Search (https://blog.google/innovation-and-ai/products/identifying-ai-generated-media-online/).
- For video, Meta asks users to disclose it, and YouTube asks creators to disclose realistic synthetic content (https://support.google.com/youtube/answer/14328491).
For a Polish team the practical split is simple: photos and video already meet visible labels on Facebook, Instagram, LinkedIn and TikTok, while text does not. Whether your channel needs its own disclosure under Article 50 is a question for a lawyer.
A note on scope: Claude's own marks cover text and, according to Anthropic's page, files such as PNG and JPEG. The page does not describe video, so the video part above is about AI video labels in general, not about Claude.
How the platforms your Polish audience uses treat AI content
For Polish readers, Facebook dominates. NapoleonCat's August 2026 estimates count about 24.8 million Facebook users in Poland (66.2% of the population), 13.5 million on Instagram (36%) and 9.9 million on LinkedIn (26.5%), plus 19.9 million on Messenger (NapoleonCat: https://stats.napoleoncat.com/social-media-users-in-poland/2026/). These are advertising-audience estimates, not a survey.
For text, what matters is that the documented labels concern media. Meta reads C2PA and IPTC markers and invisible watermarks on images and asks users to disclose video and audio (Meta: https://about.fb.com/news/2024/02/labeling-ai-generated-images-on-facebook-instagram-and-threads/), and LinkedIn shows a CR label on images and video with C2PA credentials (Fortune: https://fortune.com/2024/11/26/linkedin-content-credentials-labeling-challenges-c2pa). On text, LinkedIn's statement of July 9, 2026, cited by 404 Media (https://www.404media.co/linkedin-and-x-are-flooded-with-ai-spam-browsing-data-suggests/), says it works to "reduce low quality, automated or generic content." I found no platform that says it reads Claude's text watermark. Whether a platform label would cover your own duties under the AI Act is a question for a lawyer.
A workable routine for a team in Poland
- Let Claude draft, and let a named editor decide what is published.
- Add local facts that a model cannot supply: your own Google.pl data, your clients' results, screenshots.
- Check numbers and links before release.
- Keep a short note of who reviewed each page, since editorial responsibility is what the EU text asks about.
- Do not chase the removal of the mark. Its SEO effect is unproven, and real editing may remove it anyway.
How to measure the effect on your own Google.pl pages
Studies about other sites will not settle the question for yours, so measure it. A simple method that a small team can run:
- List every page with its publish date, and mark whether it predates August 2, 2026.
- In Search Console, filter Performance by page and by the country Poland, so traffic from other countries does not blur the picture.
- Compare the first 28 days after publishing with similar earlier pages, and not with your single best page.
- In GA4, separate visits from the Google organic source and from the AI assistant channel. Both will be small at first.
- Decide in advance what would change your mind, for example a gap of several positions across ten similar pages. That keeps a handful of visits from turning into a story.
FAQ
Does the AI Act require me to label my blog posts? It depends on the type of text and on your editorial responsibility. Ask a lawyer, because this article gives no legal advice.
Does Google.pl rank Claude content lower? No source has shown that. The documented risk is publishing low-value pages at scale.
Is Claude Code marked as well? Yes. Anthropic lists it among the products whose text is marked.
Do social platforms read the Claude watermark? I found no platform that says it reads Claude's text watermark. Documented labels concern images and video, and platforms also judge behavior, such as generic or automated posts.
Which platforms should a Polish team watch? Facebook (about 24.8 million users in Poland), Instagram (13.5 million) and LinkedIn (9.9 million) by NapoleonCat's August 2026 estimates. Their documented AI labels concern images and video.
Summary
For teams in Poland the watermark is first a compliance topic and only second an SEO topic. Nothing published links it to Google.pl rankings, and quality with human responsibility remains the safe strategy. Read more in Does AI-generated content rank in Google in 2026? and 74% of new web pages are AI-generated. To turn content into steady visibility on Google.pl, see our SEO marketing and content optimization services.
