Claude’s GEO Crackdown: Earned Media Is the Real AI Visibility

Anthropic has updated the usage policy for Claude, its AI assistant, to ban using it for the “deceptive campaigns” and “artificial activity” that some sellers of GEO (generative engine optimization) have pushed as a shortcut to AI visibility. The new rules take effect Nov. 12, and a PRmoment column called the ban “objectively very good news.”

This is good news for anyone doing real PR. For two years, “GEO” gurus have sold shortcuts to get brands into AI answers: fake reviews, planted forum posts, spun content farms. Now the AI platforms themselves are starting to shut that down.

What AI tools actually trust is earned media. Real stories in real outlets, each carrying a backlink, are what make a company visible on Google and in ChatGPT, Claude and Gemini, and they don’t get banned next quarter.

I run Prism PR, and we build our own AI tools for PR. Here is what the crackdown means for executives, founders and communications teams, and what to ask your agency on Monday.

What Anthropic’s New Rules Ban

The new section of the policy is titled “Do Not Engage in Deceptive Campaigns or Artificial Activity.” Under it, Anthropic bars using Claude to:

• Create or operate fake personas, accounts, media outlets or organizations, including “sockpuppets, astroturfing, fake reviews, or bots that claim to be human.”
• Hide who paid for content meant to influence public opinion or decision-makers.
• Push content through “networks of fake or ostensibly independent outlets, websites, or accounts to hide that it comes from a common source.”
• “Manipulate the sources from which search engines or AI systems draw answers by seeding them with content that misrepresents its origin, authorship, or independence.”
• Build the tools that make any of this possible, or “market, sell, or solicit funding for any of the above as a product or service.”

That fourth line is the one aimed squarely at the AI visibility shortcut business. Seeding the web with content that pretends to be independent, so that AI tools repeat it, is a named violation under the new rules.

Google already has similar rules on the search side. Its spam policies already treat “scaled content abuse,” pages “generated for the primary purpose of manipulating search rankings and not helping users,” as spam, and they name generative AI as one way it gets made.

To be fair about what this does and does not do: a policy is not a filter. The PRmoment column makes the same point, noting that “a usage policy update won’t make these dodgy tactics disappear totally.” People will keep trying to game the system. But the direction is clear, and a brand that builds its AI visibility on tactics the platforms have banned is building on sand.

Why Earned Media Is the AI Visibility That Lasts

Here is what the data says AI tools actually cite. Muck Rack analyzed more than 25 million links cited by ChatGPT, Claude and Gemini for its May 2026 “What Is AI Reading?” report and found that about 84% of citations come from earned media. Muck Rack’s earned group is broad: journalism, academic research, government and nonprofit sources, encyclopedic sites like Wikipedia, social platforms and third-party corporate content. Journalism alone was about 27% of all citations. Paid and advertorial content was just 0.3%.

In plain terms, most of what AI tools cite is content the brand did not publish or pay for itself. Owned and paid content together made up only about 16%.

This matters for SEO and AEO: showing up when people search on Google, and when they ask AI tools like ChatGPT, Claude and Gemini. A real story in a real outlet does both jobs. It is the kind of source AI tools cite, and it carries a backlink (a clickable link from a news story to your website, which lifts your Google ranking). We wrote more about this in how to get your brand mentioned in ChatGPT and in earned media vs. paid media.

Earned coverage is also the hardest thing to fake: a reporter chose to write it, an editor chose to run it, and the outlet’s name is on it.

Ask Your Agency How It Will Get You Cited by AI

My advice to any executive: ask your agency exactly how it plans to get you cited by AI tools. If the answer isn’t earned coverage, walk away.

Here is how to have that conversation:

• Ask for the plan in one sentence. A good answer sounds like “we will land stories about you in outlets your buyers read.” A bad answer is a pile of jargon about prompts, entities and content velocity with no reporter anywhere in it.
• Ask who publishes the content. If the answer is “we do,” on sites the agency owns or pays for, that is not earned media. Compare it with the list above: networks of outlets that only look independent are exactly what Anthropic just banned.
• Ask about reviews and forums. Any plan to post reviews, Reddit threads or forum answers that look like they come from real customers but are written for you is a fake review or a sockpuppet by another name.
• Ask to see the coverage. Real earned media has a byline, an outlet and a link you can click. Ask for examples from the agency’s own clients.
• Ask what happens when the rules change again. Tactics that depend on a loophole end when the loophole closes. Stories in real outlets keep working.

How We Use AI at Prism: More Real PR, Not Fake PR

At Prism we build our own AI tools for PR: media monitoring, newsjacking and fact checking. So we use AI to do more real PR, not to fake it.

That is the line I would draw for any company. Use AI to watch the news faster, find the stories your experts can speak to, and check every fact before it goes to a reporter. Do not use it to invent people, reviews or outlets.

Prism is an award-winning, full-service PR agency, and we are also a digital PR and linkbuilding firm. Our PR services and our earned media for SEO and AEO run on one idea: earn real stories, each carrying a backlink. One way in is a newsworthy, data-driven industry study; our 2026 industry study campaigns average 20.8 stories a month. Company-wide, we have placed 25,000 stories and earned 20,000 backlinks.

Frequently Asked Questions

What did Anthropic change in Claude's usage policy?

Anthropic added a section called “Do Not Engage in Deceptive Campaigns or Artificial Activity.” It bars using Claude to create fake personas, outlets or reviews, hide sponsorship, run networks of fake independent sites, or seed the sources search engines and AI systems draw on with content that misrepresents where it came from. The updated policy takes effect Nov. 12, 2026.

Does the policy ban GEO?

Not by name: the policy never mentions GEO, and it does not ban trying to show up in AI answers. It bans the deceptive ways of doing it: fake sources, hidden sponsorship and content networks that pretend to be independent. Earning real coverage in real outlets is untouched.

What is the best way to improve AI visibility?

Earn coverage. Muck Rack’s May 2026 report found about 84% of AI citations come from earned media, a group that includes journalism, academic and government sources, encyclopedic sites and social platforms. Stories in real outlets, with backlinks, help you on Google and in AI answers at the same time.

What should I ask a GEO or AI visibility agency?

Ask exactly how it plans to get you cited by AI tools, who publishes the content, and to see real coverage it has earned for clients. If the answer is not earned coverage, walk away.

Build AI Visibility That Lasts

The AI platforms are starting to shut the shortcuts down. The companies that come out ahead will be the ones that already have real stories in real outlets.

If you want AI visibility that does not get banned next quarter, talk to us.