--- title: "Claude's Usage Policy Now Bars Seeding Search and AI Answers Through Fake-Independent Sites" description: "From 12 November, Claude's Usage Policy bars swaying search and AI answers through sites posing as independent, and drops publishing from its high-risk list." tldr: "Anthropic's new Usage Policy, effective 12 November 2026, prohibits using Claude to manipulate the sources search engines or AI systems answer from by seeding them with content that misrepresents its origin, authorship or independence, such as networks of sites posing as unaffiliated. It covers API customers and the end users of products built on Claude, and the penalty is restricted or lost access. The same update drops automated content publishing from the high-risk list that required human review and AI disclosure. The rewritten high-risk rules apply to recommendations about a specific person and exclude general educational content." publishDate: 2026-10-12 author: "Liam Hayward" --- Anthropic has published a new version of its Usage Policy, the rules that apply to anyone using Claude, whether through its apps, its API, a cloud provider or a third-party product built on it. It takes effect on 12 November 2026. Two changes bear on SEO work: a new prohibition aimed at manipulating the sources that search engines and AI systems draw answers from, and the removal of automated publishing from the list of high-risk uses that carry review and disclosure requirements. Anthropic's announcement post of 8 October names neither. Both show up only when the new text is read against the September 2025 version, a comparison Search Engine Journal published on 9 October. ## What does Claude's new deceptive campaigns section prohibit? The policy has a new section, "Do Not Engage in Deceptive Campaigns or Artificial Activity". One of its lines is written about search: > Manipulate the sources from which search engines or AI systems draw answers by seeding them with content that misrepresents its origin, authorship, or independence (e.g., networks of sites posing as unaffiliated sources corroborating the same claims) The lines around it bar distributing content "through networks of fake or ostensibly independent outlets, websites, or accounts to hide that it comes from a common source", creating fake personas, outlets or reviews to mislead people about a content's origin, building tools designed to run such campaigns, and selling any of it as a service. The September 2025 policy did not mention search engines or AI answers anywhere, though it already prohibited spam and fake personas. Anthropic's post says the policy already prohibited fake-account networks and fabricated news sites, but that the rules were scattered across its elections, fraud, privacy and disinformation sections, and that the consolidated section applies to deceptive activity "whether political or commercial". Enforcement is against the user: Anthropic says it may warn, throttle, limit, suspend or terminate access. The post points to Anthropic's September threat intelligence report for the pattern behind the section. One case in it is a commercial "influence-as-a-service" operation that Anthropic traced to LKM Company, a France-based digital advertising agency: about 70 fabricated news sites made to look like independent local newsrooms, publishing political stories slanted for whoever was paying. Claude ran in an automated pipeline that produced articles with three to four internal links each, signed with the names of journalists who did not exist, and the network published at least 8,913 of them in about 20 languages. In the report's words, the articles were designed "to boost their site's authority rankings on search engines". Anthropic banned the account and the organisation. The report says most of the content drew little engagement from real audiences, and does not say whether the sites ranked or were cited in AI answers. ## Does publishing Claude-written content still count as high-risk? Not under the new text. The September 2025 policy listed "Media or professional journalistic content", defined as using Anthropic's products "to automatically generate content and publish it for external consumption", among its high-risk use cases. Those carried two requirements: a qualified professional had to review advice or recommendations before they were disseminated, and AI involvement had to be disclosed when outputs were presented directly to people. The new version applies the same two controls to "High-risk AI Recommendations" in 11 named areas: legal, medical, finance, credit, insurance, housing, employment, education and credentials, healthcare access, public benefits, and legal status. Publishing is not among them. Anthropic's post says the high-risk requirements "haven't changed" and that the section was rewritten to state more directly which uses are covered, but it does not mention publishing or say why the category went. The new text does answer the question a publisher on a medical or financial topic would ask next. The controls apply to a recommendation made to or about an individual, and the exclusions name "general information or educational content, including explaining what a law says, how a condition or treatment works, or how investing works, without applying it to the individual's circumstances". A published explainer falls on the excluded side; a tool that tells a specific person which treatment or fund to choose does not. Two rules that matter for content carry over unchanged in substance: Claude may not be used to "submit AI-assisted work without proper permission or attribution", and fake personas used to mislead people about a piece's origin or attribution remain prohibited, wording that on its face takes in an invented author byline. ## What the Claude policy change means for SEO work If you, an agency or a content tool you pay for uses Claude to write for several sites presented as unconnected, such as a private blog network or a set of "independent" review sites backing the same product, that is a named violation from 12 November. The old bans on spam and fake personas arguably covered it already; the new line removes the argument. The cost is losing access to Claude, which is separate from anything Google's spam policies do to the sites themselves. The policy reaches the end users of products built on Claude, so ask the agencies and content tools you pay which model they run on and whether their output feeds sites like these. For ordinary publishing, Anthropic's policy no longer requires its review-and-disclose steps for Claude-written articles as such. That is Anthropic's rule only: it does not change how Google treats AI-generated content, and the policy says nothing about Anthropic's separate commitment to [watermark all Claude-generated text](/seo-news-updates/claude-marks-ai-generated-content/). The policy governs what people do with Claude. It says nothing about how Claude selects or cites sources when it searches the web, so it is not a signal about Claude's retrieval. ## Sources - [2026 Usage Policy update, Anthropic, 8 October 2026](https://www.anthropic.com/news/2026-usage-policy-update) - [Usage Policy, effective 12 November 2026, Anthropic](https://www.anthropic.com/legal/aup) - [Usage Policy, effective 15 September 2025 (archived), Anthropic](https://www.anthropic.com/legal/archive/22742366-2ef0-4c7a-a833-6523f10d3944) - [Threat intelligence report, September 2026, Anthropic](https://www.anthropic.com/threat-intelligence-report-september-2026) - [Claude's New Rules Target Fake Sources Built To Sway AI Answers, Search Engine Journal, 9 October 2026](https://www.searchenginejournal.com/claude-rules-fake-sources-ai-answers/592456/)