Will Claude's New Rules Stop Fake Sources From Skewing AI Answers?

Not by itself. Anthropic's updated usage policy, effective Nov. 12, bars people from using Claude to seed search and AI sources with content that hides its origin, authorship, or independence. It only binds Claude users, and the reporting so far does not show it changing any AI answer. Your real, verifiable listings still do the work.

Anthropic released the update on Oct. 8, according to Matt G. Southern at Search Engine Journal. Today is Oct. 12, so the new section is not even in force yet. Anyone telling you it is already reshaping AI citations is guessing.

Still, it matters to a contractor. Here is what the policy says, what Anthropic found, and what I would do about it.

What does Anthropic's new Claude policy actually say?

The policy adds a section titled "Do Not Engage in Deceptive Campaigns or Artificial Activity." It forbids using Claude to manipulate the sources search engines or AI systems draw answers from, by seeding them with content that misrepresents where it came from, who wrote it, or whether it is independent.

Southern reports that the policy's own example is a network of sites posing as unrelated sources that back the same claims. In the policy's words: "networks of sites posing as unaffiliated sources corroborating the same claims."

Other lines in the section bar fake personas, fake outlets, and fake reviews. They also bar building tools meant for deceptive campaigns and selling any of this as a service.

The earlier version, effective Sept. 15, 2025, did not mention search engines or AI responses. Southern is careful to say that "not named" does not mean "allowed." Fake reviews were prohibited in both versions.

The rules apply to everyone using Anthropic's products, from developers on the API to people using apps built on Claude. If Anthropic suspects a violation, it may suspend or terminate access.

What did Anthropic find that led to this rule?

In a September threat report, Anthropic described around 70 websites that looked like independent local news outlets but were all tied to one France-based ad agency. Claude was used to write articles in a consistent format for automated publishing, built to boost the sites' authority in search rankings.

According to Southern's summary of that report, the network published at least 8,913 articles across around 20 languages. Most got little engagement from real readers.

The report does not say whether the sites actually improved their rankings or showed up in AI-generated answers. That is worth remembering. The spam was built at scale, but nobody has shown it worked.

Anthropic removed the Claude account and banned the organization behind it.

You have probably seen the contractor version of this play. A pile of "local" sites all declare the same company the best in town. None of them is a real review, and none is independent. That is exactly the pattern the policy describes.

Will fake directories and review sites disappear from AI answers?

No, not because of this policy alone. It covers Anthropic's customers, which means content made with Claude. Spam written by hand or with other tools is not covered, and other AI companies set their own rules. Treat this as a signal about direction, not a cleanup.

The enforcement tool here is account suspension. That stops a bad actor from producing more content through Claude. It does not delete the junk that already exists on the web.

So if a fake directory is sitting in your category today, assume it can still be there in January. Hoping the garbage clears out is not a plan.

The useful part is the direction. An AI company is writing "fake independence" into its rules. If you have ever been tempted by a cheap package of fake reviews or a network of look-alike sites, the risk is going up, not down.

What does this mean for a home-services business?

Two things. Do not buy fake reviews or build a web of look-alike "local" sites, because the rules now name that behavior. And do not count on spam leaving your category. The safe bet is a profile and a set of listings that are accurate, consistent, and real.

Reputation is where this hits hardest. Fake reviews were banned in both versions of the policy, and the new section adds fake personas and fake outlets on top. Real reviews from real customers are the one asset nobody can accuse you of faking.

Say a competitor in your town pads their profile with reviews from people who never hired them. That is bad for you, but it is also the kind of thing platforms are being pushed to police. Document it. Report it through the platform's own tools. Do not answer with fake reviews of your own.

Your own profile is the other half. If your Google Business Profile has wrong hours, a missing category, or a name that does not match your other listings, you are giving an AI assistant nothing solid to work with. Our Google Profile Audit shows you where the gaps are.

How do I build citations AI can trust?

Make every place your business appears say the same true thing, and let real customers do the talking. That means a complete Google Business Profile, matching name, address, and phone everywhere, and reviews you earn after real jobs. It is boring work, and boring work holds up.

Here is the process I would follow:

1. Finish your Google Business Profile. Add every service category, your service areas, hours, real photos of real jobs, and a clear description.
2. Match your name, address, and phone number everywhere. Check your website, directories, and social pages. One small mismatch can make you look like two businesses.
3. Ask every happy customer for a review right after the job. Send the link while they are still glad you showed up. Reply to every review, good or bad.
4. Check what AI says about you. Ask ChatGPT, Perplexity, and Google for the best business in your trade in your town. Note which sources they name and whether you appear. Our AI Visibility Report does this for you.
5. Report fake listings and fake reviews. Use each platform's reporting tool, keep screenshots, and move on. Do not fight spam with spam.

If you are not sure where you stand, the free Local Visibility Score is a quick place to start.

The shortcut was always a bet that nobody would check. More and more, the tools do check. A real customer who left a real review is the one thing a fake network cannot copy.

Common questions

Does the Claude policy apply to my business?

Only if you use Claude, or an app built on it, to produce content. Southern reports the rules cover developers on the API and everyday users of apps built on Claude. If you are not faking reviews, authorship, or independence, nothing about how you run your business has to change.

When does the new Claude usage policy start?

Anthropic released it on Oct. 8, and it takes effect Nov. 12. The previous version took effect Sept. 15, 2025. Until Nov. 12 the new section is not in force, though fake reviews were already prohibited in both versions. Anyone claiming it has already changed AI answers is guessing.

Is it okay to use AI to help write my Google Business Profile posts?

The new section targets deception about origin, authorship, and independence. It does not target using AI as a writing tool. Posts about your real jobs, under your real business name, are not what it describes. Keep them accurate, and never pretend to be someone you are not. Read the policy yourself if unsure.

Will AI assistants stop recommending spam directories?

Nothing in the source shows that. Southern notes that Anthropic's report does not say whether the fake news network improved rankings or appeared in AI answers. Plan on spam sticking around for a while, and make your own listings and reviews strong enough to stand out next to it.