On October 8, Anthropic updated its Usage Policy for Claude. The new version takes effect on November 12, 2026. For affiliates and SEO teams, one change stands out: Claude is now explicitly off-limits for building site networks and fake "independent" sources designed to game search results and AI answers.
What changed in the policy
The policy adds a dedicated section, "Do Not Engage in Deceptive Campaigns or Artificial Activity," covering both political and commercial deception. Similar rules used to be scattered across the elections, fraud, privacy and disinformation sections. Now they sit in one place, with a wider scope.
The key new rule: you can't use Claude to manipulate the sources search engines and AI systems draw answers from by seeding them with content that hides its origin, authorship or independence. Anthropic's example is networks of sites posing as unaffiliated sources that back up the same claims. Fake personas, fake outlets and fake reviews are banned too, along with amplification through sock-puppet accounts, building tools for such campaigns and selling them as a service.
Where the ban came from: 70 "local news" sites
The trigger was a case from Anthropic's September threat report. Around 70 websites posed as independent local news outlets and, according to Anthropic, were tied to a France-based ad agency. Claude was used to write articles in a fixed template with 3–4 internal links each, published automatically: at least 8,913 articles in about 20 languages. The goal was to boost the sites' authority in search. Most of the articles drew almost no real readers, and Anthropic banned both the account and the organization behind it.
What it means for affiliates and SEO
Writing content with Claude is still fine. What the rule targets is disguise: content presented as independent when one owner is behind it. In affiliate marketing, the risk zone includes:
- PBNs and networks of "independent" sites that link to each other and repeat the same claims;
- fake review sites and rankings for casinos, nutra offers and affiliate programs written by made-up authors;
- fabricated reviews and fake "experts" on forums and social media;
- GEO manipulation: mass-seeding content so ChatGPT, Gemini and other AI assistants recommend a specific brand.
Openly attributed content, such as your own site, blog or reviews with a clear author, isn't covered by the rule as worded. Anthropic also dropped automated content publishing from its list of high-risk use cases, so generating and publishing content is generally fine as long as it doesn't pose as an independent source.
Violations can lead to restricted or terminated access to Claude, including through the API and third-party apps built on it. Keep Google in mind too: its spam policies already treat attempts to manipulate AI responses in Search as spam. The difference is that Google penalizes websites, while Anthropic acts against users.
Other changes
- The blanket ban on personalized political campaign targeting is lifted, but deceiving voters remains prohibited.
- "Sustained and needless" cruelty toward the model is now prohibited. This only covers extreme cases; ordinary pushback and criticism don't count.
- Rules on surveillance, weapons and autonomous hardware are stricter.
If you generate content for site networks with Claude, check before November 12 that none of your sites pose as independent sources. Otherwise you risk losing access to the service.