Anthropic bars using Claude to seed fake AI answer sources

Anthropic published an updated Usage Policy on 8 October 2026, with an announcement summarizing what changed. The policy page is marked effective 12 November 2026. Matt G. Southern of Search Engine Journal reported the search-related line the same week. We read both Anthropic pages directly, and the wording quoted below is theirs.

What does the new rule say?#

The line sits under a new heading, “Do Not Engage in Deceptive Campaigns or Artificial Activity”. Among the prohibited uses of Claude, it names manipulating the sources from which search engines or AI systems draw answers by seeding them with content that misrepresents its origin, authorship or independence. The policy’s example is networks of sites posing as unaffiliated sources corroborating the same claims.

That makes it a rule about AI search inputs, and a close cousin of the link building abuse of fake independence: many sites that look unconnected but share one source.

What else is in the same section?#

Per the policy page, the same section also bars:

• Creating or operating fake personas, accounts, media outlets or organizations to mislead people about origin, attribution or how widely a view is held, including fake reviews.
• Concealing the sponsorship of content meant to influence public opinion or officials.
• Distributing content through networks of fake or ostensibly independent outlets, websites or accounts to hide that it comes from a common source.
• Building, improving or maintaining tools or infrastructure designed to enable deceptive campaigns.
• Marketing or selling any of the above as a product or service.

Anthropic’s announcement says its old policy already prohibited this activity, but the restrictions were scattered across the elections, fraud, privacy and disinformation sections. The new section consolidates them and, it says, applies to deceptive activity of any kind, political or commercial.

What does it not cover?#

Three limits are clear from the sources. First, it is a usage policy for one vendor’s products. Anthropic’s policy says its Safeguards Team enforces it, and that if it suspects a violation it may warn the user or throttle, limit, suspend or terminate access. Southern contrasts this with Google, which enforces its spam policies on websites, where a site can rank lower or disappear from results.

Second, it is not a search engine rule. Southern notes the announcement describes the new section but does not mention search engines. Southern also says Google’s spam policies already treat attempting to manipulate generative AI responses in Google Search as spam, which is a separate rule from a separate company.

Third, neither Anthropic page says how any of this is detected, by Anthropic or by a search engine, so this article makes no claim about it. For how Google frames scaled content, see Google ties more spam updates to a flood of new content.

What did Anthropic say prompted it?#

The announcement points to examples Anthropic wrote about the month before: state media outlets, government propaganda offices and commercial firms using Claude to run networks of fake accounts and fabricated news sites.

Southern adds detail from that September threat report, which we did not read. As he reports it, Anthropic identified around 70 websites posing as independent local news outlets, all linked to a France-based ad agency, that used Claude to produce articles with consistent formatting and internal links for automated publishing. He says the report did not state whether the sites gained rankings or appeared in AI answers, and that Anthropic removed the account and banned the organization.

What changed for publishing?#

Southern reports a second change. The earlier version, effective 15 September 2025, listed “Media or professional journalistic content”, described as automatically generating content and publishing it for external consumption, as a high-risk use needing human review. He says the new version lists 11 high-risk areas and publishing is not among them. We checked the current policy page: its 11 areas are legal, medical, finance, credit, insurance, housing, employment, education, healthcare access, public benefits and legal status, and none is publishing.

Southern stresses this is a change to the high-risk list, not a blanket exemption. He adds that Anthropic’s announcement says its high-risk requirements have not changed and does not explain the removal. For how Google treats AI-written pages, see Google says fact-checking AI content is critical before publishing.

What should publishers check?#

Southern’s advice is to check whether content made with Claude, by you or by an agency, is published across a network of sites presented as independent of each other. That is the verifiable part: ask your agency which sites it publishes to and who owns them. Anything that presents a shared source as independent is the pattern the policy names, whatever tool wrote it.