Case study: Gumlet turned ChatGPT mentions into 20% of inbound revenue. Read it →
Claude Bans Fake “Independent” Sources for AI Answers: What GEO Buyers Should Check Before Nov 12 (Oct 2026)
TL;DR
- Anthropic’s 2026 Usage Policy update, published Oct 8 and effective November 12, 2026, adds a section called “Do Not Engage in Deceptive Campaigns or Artificial Activity.”
- One line in the new Usage Policy bans using Claude to “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 same section also bans selling any of this as a service: “Market, sell, or solicit funding for any of the above as a product or service.”
- Google already treats this as spam. Its spam policies define spam to include “attempting to manipulate generative AI responses in Google Search,” and the September 2026 spam update finished rolling out on Oct 8.
- For B2B SaaS buyers, the practical question is simple: is any of the third-party coverage your GEO agency produces for you secretly coming from one source? Ask before November 12, and write disclosure into the contract.
Most B2B SaaS teams buying GEO in 2026 are paying for one thing above all: getting named when a buyer asks ChatGPT, Claude, Perplexity or Google AI Mode for a shortlist. Those answers lean on third-party pages, such as listicles, review roundups and comparison posts. That makes third-party coverage the most valuable asset in AI search, and the easiest one to fake.
On October 8, the company behind Claude wrote down, in plain language, that faking it with Claude is off limits. Google had already said the same thing about its own AI answers. Two of the largest answer engines now describe the same pattern as manipulation. If your AI visibility depends on that pattern, the risk is no longer theoretical, and it sits with you, not just with your agency.
This piece covers what actually changed, what it does and does not cover, and a short checklist to run with your agency before November 12.
What happened
On Oct 8, Anthropic published its annual Usage Policy update. The post says the new version “takes effect on November 12.” Its headline change on deception reads:
“We’ve consolidated those rules into a new section titled Do Not Engage in Deceptive Campaigns or Artificial Activity, which applies to deceptive activity of any kind (whether political or commercial).”
The announcement itself does not mention search engines. The search language is in the policy text, which Search Engine Journal’s Matt G. Southern flagged on Oct 9. The new section lists six prohibited uses. Four of them map directly onto tactics sold under the GEO label:
| Policy line (quoted) | What it looks like in GEO |
|---|---|
| “Create or operate fake personas, accounts, media outlets, or organizations … (e.g., sockpuppets, astroturfing, fake reviews, or bots that claim to be human)” | Invented “editors” on review sites, fake Reddit or G2 voices |
| “Distribute content through networks of fake or ostensibly independent outlets, websites, or accounts to hide that it comes from a common source” | One owner running several “independent” software review blogs |
| “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)” | Ten listicles on ten domains that all rank the client #1, with no stated link to the client or agency |
| “Market, sell, or solicit funding for any of the above as a product or service” | Agencies packaging the above as a “citation network” or “AI mention” product |
Two more lines matter for buyers: “Misrepresent or intentionally conceal the sponsorship of content or messaging intended to influence public opinion, officials, or other decision-makers,” and building “tools, systems, or infrastructure designed to facilitate or enable deceptive campaigns.”
Enforcement is about access. The policy says that if Anthropic suspects a violation, “we may warn you or throttle, limit, suspend, or terminate your access to our products and services.” It applies to “anyone who submits inputs to Anthropic’s products or services,” including API developers and users of apps built on Claude.
Anthropic ties the change to real cases. Per SEJ’s reading of Anthropic’s September threat intelligence report, Anthropic found around 70 sites that looked like independent local news outlets, all linked to one France-based ad agency. They published at least 8,913 articles in about 20 languages, built with internal links that the report says were “specifically designed to boost their site’s authority rankings on search engines.” Anthropic banned the account and the organization.
Google already calls this spam
Google’s spam policies page opens with this definition:
“In the context of Google Search, spam refers to techniques used to deceive users or manipulate our Search systems into featuring content prominently, such as attempting to manipulate Search systems into ranking content highly or attempting to manipulate generative AI responses in Google Search.”
SEJ reports that this wording came from a May 15 documentation update clarifying that the policies apply to generative AI responses in Search. Several existing policies on that page describe the network pattern directly:
- Scaled content abuse includes “Creating multiple sites with the intent of hiding the scaled nature of the content.”
- Doorway abuse includes “Having multiple websites with slight variations to the URL and home page to maximize their reach for any specific query.”
- Thin affiliation covers programs that distribute content “across a network of affiliates without providing additional value.”
- Policy circumvention covers “Using existing or creating new subdomains, subdirectories, or sites with the intention of continuing to violate our policies.”
Google’s September 2026 spam update started Sept 24 and finished Oct 8. SEJ notes that Google did not say what it targeted, so don’t read it as an “AI answer spam” update. What matters is that the policy wording is now in force and Google enforces it on sites: violating sites “may rank lower in results or not appear in results at all.”
What the change covers, and what it doesn’t
It is easy to over-read this, so be precise.
It does not ban AI-written content. The rule is about misrepresenting origin, authorship, or independence. SEJ also points out that the new policy no longer lists automatically generating and publishing content as a high-risk use case. Google’s own guidance has long said AI use is not the problem; low-value content made to manipulate rankings is.
It does not ban owning media. Nothing in the text stops a company or an agency from running a publication. The line is drawn at hiding the connection. A site that clearly says who owns it and how it relates to the companies it covers is not “posing as” an unaffiliated source.
It does not reach content Claude didn’t touch. Anthropic can only cut off its own users. A network written with another model, or by hand, is outside Anthropic’s policy, though still inside Google’s spam policies if it targets Search.
It does not remove anything from AI answers directly. Anthropic enforces against accounts, not web pages. The near-term risk for a buyer is less “ChatGPT stops citing us tomorrow” and more: your agency’s production stack loses its main tool, Google demotes the sites carrying your mentions, and your AI visibility built on those sites disappears with them.
It does add a commercial risk. The policy now names selling these tactics “as a product or service.” An agency whose GEO offer is mostly undisclosed placements has a business model that two major platforms describe as prohibited.
Clean GEO: earned, disclosed, verifiable
AI answers favor brands that many independent sources describe the same way. That is exactly why fake independence is tempting, and exactly why it is now named. The durable version of the same goal looks like this:
- Earned coverage. Real reviews, analyst mentions, customer case studies on partner sites, podcasts, and journalists citing your data. The publisher chose to cover you.
- Disclosed coverage. Sponsored placements, affiliate listicles and sister publications are fine when the relationship is stated on the page. For paid links, Google’s policies say sponsorship is acceptable “as long as they are qualified with a rel=”nofollow” or rel=”sponsored” attribute value.”
- Verifiable claims. Pages that answer engines cite should carry checkable facts: pricing, integrations, named customers, and original data. Corroboration from honest sources is worth more than ten echoes of one source.
A note on our own setup, since you should ask any agency the same thing: DerivateX has affiliated media properties, and they disclose their relationship to DerivateX. We think every agency should be able to say that sentence, with a list.
Your pre-November 12 checklist
These are questions for your current or prospective GEO agency. For the broader vetting list (pitch, proposal, contract and first 90 days), use our 27 SaaS SEO and GEO agency red flags. The items below cover only the source-manipulation risk.
1. Get the full list of where you’ve been placed.
Ask for every third-party URL the agency produced, paid for or arranged that mentions your brand, with the date and how each one came about (earned, paid, affiliate, owned).
2. Ask the ownership question directly.
“Do you, your staff, or any partner own, operate or control any site, review platform, newsletter, or social account where our brand has been or will be mentioned?” Get the answer in writing.
3. Check for the network fingerprint yourself.
Open five of the listicles that rank you. Look for the same template, the same “editorial team” with no real bios, identical rankings, overlapping WHOIS or hosting, and cross-links between them. One or two coincidences are normal. A pattern is not.
4. Ask what each placement says about the relationship.
Every page with a commercial or ownership link to you or your agency should say so on the page, not in a footer elsewhere.
5. Ask which AI tools they use to produce third-party content.
If the answer includes Claude, the November 12 policy applies to that production directly. If you build with the Claude API yourself, the policy applies to you too.
6. Put disclosure in the contract.
Add three clauses: the agency will not create or use undisclosed affiliated sources, fake personas or fake reviews on your behalf; the agency will disclose any ownership or financial relationship with any site where it places your brand; and you can terminate if either is breached.
7. Re-baseline your AI visibility on clean sources only.
Measure which AI answers cite you, then remove anything coming from sources you’d be uncomfortable defending. What’s left is your real starting point.
How DerivateX thinks about it
Our GEO work treats third-party corroboration as one leg of citation, alongside owned content that is easy to extract and consistent entity signals. The new policy doesn’t change that model. It removes the shortcut. If your current AI visibility can’t survive questions 1 to 4 above, it’s worth knowing now rather than after an enforcement wave.
If you want to see where you’re actually cited and what’s holding those citations up, start with a free AI visibility audit, or see pricing for the 90-day GEO pilot.
FAQ
When does Anthropic’s new rule take effect?
November 12, 2026. It was published on October 8, 2026.
Does the rule ban using Claude to write blog posts or listicles?
No. It bans content that “misrepresents its origin, authorship, or independence.” Clearly attributed content is not covered by that line.
Is a sponsored listicle now against the rules?
Not if the sponsorship is disclosed. The policy prohibits concealing sponsorship, and Google allows paid links that use rel=”sponsored” or rel=”nofollow.”
Can an agency own a media site that covers its clients?
The text targets sites “posing as unaffiliated sources.” A clearly disclosed affiliation is a different thing. Hiding it is the problem.
What happens if an agency breaks the policy?
Anthropic says it may “warn you or throttle, limit, suspend, or terminate your access.” Separately, Google says sites violating its spam policies “may rank lower in results or not appear in results at all.”
Was Google’s September 2026 spam update aimed at AI answer manipulation?
Google hasn’t said what it targeted. It ran Sept 24 to Oct 8. The policy language on generative AI responses was already in place before it.
Do OpenAI or Perplexity have the same rule?
This piece covers only Anthropic and Google, whose texts we verified. Check each provider’s current usage policy directly.
Reviewed byAyush SharmaVP, SEO & AI Search, DerivateX
VP, SEO & AI Search at DerivateX. We're a B2B SaaS SEO and Generative Engine Optimization agency that engineers AI citations in ChatGPT, Perplexity, Claude, and Gemini and connects them to demo bookings and revenue pipeline.
Cited by AI
