Key Takeaways
Anthropic published its 2026 Usage Policy update on 8 October, and it takes effect on 12 November
The policy bans using Claude to seed AI answer sources with content that fakes its origin or independence
Anthropic's example is a network of sites posing as unaffiliated sources that back the same claims
A breach can lead to a warning, throttling, suspension or termination of Claude access
Google's AI search guide already says inauthentic mentions don't help in AI Overviews or AI Mode
From 12 November, you can't use Claude to fake the sources AI answers are built from. Anthropic added the rule in its 2026 Usage Policy update on 8 October.
The Anthropic usage policy now names search engines and AI systems in writing. If your AI visibility work runs through a group of sites with one owner and separate brands, you have 27 days to unwind it.
What does the Anthropic usage policy say about AI answers?
Anthropic says the policy already prohibited this activity, but the rules sat across its elections, fraud, privacy and disinformation sections. The update pulls them into 1 section.
"Applies to deceptive activity of any kind (whether political or commercial)" - Anthropic, 2026 Usage Policy update
The full section bans using Claude to:
- Run fake personas, accounts, outlets or organisations, including sockpuppets, astroturfing and fake reviews
- Hide who sponsors content meant to sway public opinion, officials or decision-makers
- Push content through ostensibly independent sites or accounts to hide a common source
- Seed search and AI answer sources with content that fakes its origin, authorship or independence
- Build tools or infrastructure for astroturfing, sockpuppet accounts or coordinated inauthentic behaviour
- Market, sell or raise money for any of the above as a service
If you run an agency, read the last bullet twice. Selling a source network as a GEO package is a violation in its own right.
A real seeded network ran about 70 fake news sites
Anthropic's September 2026 threat report describes one it shut down. A France-based ad agency, LKM Company, used Claude to run about 70 fabricated news sites.
The network published at least 8,913 articles in about 20 languages. Its prompts asked for a fixed format and 3 to 4 internal links per article.
Anthropic says the articles were built to lift the sites' authority in search engines. Fake bylines made each site look like its own local newsroom.
Most of it drew little real engagement before Anthropic banned the account. That's the tactic the new rule now names.
Anthropic vs Google on fake sources
| Category | Anthropic usage policy | Google Search |
|---|---|---|
| What's covered | Using Claude to seed search engines or AI answer sources with content that fakes its origin, authorship or independence. Anthropic's example is a network of sites posing as unaffiliated sources that back the same claims. The same section bans fake reviews, hidden sponsorship and selling these tactics as a service. | Seeking inauthentic mentions across blogs, videos and forums. Making separate pages for every query variation to manipulate rankings or AI responses also counts, under Google's scaled content abuse spam policy. |
| Who it applies to | Anyone who sends inputs to Anthropic products: Claude.ai, Claude Code, the API, cloud providers, resellers and apps built on Claude. It covers content made with Claude, wherever it gets published. | Any site in Google Search, however the content was made. AI-assisted content is fine if it meets Search Essentials and the spam policies. |
| What gets hit | our Claude account. Anthropic's Safeguards Team runs detection and monitoring, and its threat team bans accounts it ties to influence operations. | Your pages in Search, AI Overviews and AI Mode. Google's AI features rely on both its core ranking systems and the systems that block spam. |
| What can happen | A warning, throttling, limits, suspension or termination of access. In the LKM Company case, about 70 fake news sites and 8,913 articles ended in a banned account. | Google says inauthentic mentions help less than they seem to. Its spam systems block abusive content, and its AI features depend on those systems. |
| In force | From 12 November 2026, 35 days after Anthropic published it on 8 October. Fake account networks and fabricated news sites were already banned under the old wording. | Already live. Google's AI search guide was last updated on 10 July 2026. |
The 2 rules hit the same tactic from both ends. Google filters the network out of its answers, and Anthropic can cut off the tool that built it.
What you can still do with Claude for content
The test is origin, authorship and independence. Content you publish under your own brand, on your own site, sits outside it.
The policy says it governs how Claude is used, and no industry is banned outright. If you ghostwrite for clients, check the separate line on submitting AI-assisted work without permission or attribution.
Earned citations still work. We broke down how AI systems choose which sources to cite earlier this year.
3 things to check before 12 November
Audit every client site group that has 1 owner and separate brands. Those are the sites the example describes.
Check sponsored placements for clear disclosure. Hidden sponsorship of influence content is on the same list.
Send the full usage policy to whoever runs content on the Claude API. It covers API users and resellers too.
Disclaimer: This article is AI-assisted and may contain errors, so please verify key details against the linked sources.
.png&w=3840&q=75)
