What changed in Claude's rules
Anthropic has updated Claude's usage policy, and one key change directly affects marketers and growth specialists. For the first time, the company explicitly bans creating and seeding misleading sources — pages and posts whose sole purpose is to enter an AI model's retrieval context and steer its answers.
Why it matters for AI SEO and GEO
The mechanism is simple: if a model borrows from a source and that source shapes the output, you can influence the answer by publishing content tailored to appear in context. This field already has names — AI SEO, answer engine optimization, generative engine optimization (GEO). Previously a grey area, it is now explicitly framed as a violation.
At the same time, Anthropic removed automated publishing from its high-risk category, signaling that mass automation itself isn't the main threat — deliberate misinformation targeting AI systems is.
Practical implications
- A clear precedent: manipulating sources for AI answers is a policy violation, not just an ethical question.
- Other providers (OpenAI, Google, Microsoft) will likely follow.
- Affiliates and SEO teams should rethink strategies built purely on seeding bait sources for LLMs.
Context for global and RU-speaking markets
Publishers and affiliates across regions are experimenting with GEO: buying brand mentions in AI-crawler-indexed content, writing 'quotable' articles, and engineering sources. Without a public detection mechanism, the wording still gives Anthropic legal grounds for API limits and blocks if abuse is proven.
Editorial take
Anthropic is setting a precedent the market will follow: AI usage rules increasingly overlap with search-result rules. The grey zone of AI SEO is narrowing. Strategies based solely on seeding sources for model answers become risky. The winners will be those building real reputation and transparent data — not fake sources engineered for AI output.