Claude's New Rules on Fake Sources: What Local Pros Should Know

Claude's updated rules target fake sources built to sway AI answers. Here is what that means for HVAC, plumbing, pest control, and roofing businesses online.

Published November 7, 2026 by the Digital Marketing Raleigh team.

Claude's New Rules on Fake Sources: What Local Pros Should Know

Listen to this article:

For the last couple of years, a growing number of marketers have chased a new prize: getting their clients named inside AI chatbot answers. When a homeowner asks an assistant "who is the best emergency plumber near me" or "is this pest control company legit," the business that shows up in that answer can win the call before a search results page ever loads. Wherever there is a prize, someone will try to game it. Search Engine Journal recently reported, in a piece by Matt G. Southern, that Anthropic has updated the rules for its Claude models to target fake sources built specifically to sway AI answers.

If you own a local service business, you probably did not set out to manipulate a chatbot. But you may have hired someone who promised "AI visibility," or you may have been pitched tactics that sound clever and turn out to be exactly what these rules are aimed at. This article explains what the change means in plain terms, why AI companies care about it, and what a pest control, HVAC, plumbing, or roofing company should do differently starting this month. We will keep the discussion grounded in general reasoning rather than speculation about internal details we cannot verify, and we encourage you to read the original reporting and Anthropic's own published materials for the specifics.

What the News Says, and What It Does Not

The core of the report is simple: Claude's guidelines now address sources that exist mainly to influence what an AI system says, rather than to inform people. Think of pages, sites, or content networks created so that a language model will pick them up, treat them as independent evidence, and repeat their claims. The direction is clear even if you never read a line of the underlying policy. AI developers want their assistants to weigh information the way a careful researcher would, and a careful researcher discounts material that appears manufactured to produce a particular conclusion.

It is worth being precise about what this does not mean. It does not mean AI assistants will stop recommending businesses. It does not mean marketing content is forbidden or that a company website will be ignored. A business describing its own services on its own site is not a fake source; it is a first-party source, and any reasonable reader understands it as such. The concern is with content that pretends to be something it is not: a "neutral" review roundup secretly written by the company it ranks, a network of look-alike sites that all cite one another, or pages stuffed with instructions meant for machines rather than people.

We also want to be honest about the limits of what anyone outside an AI company knows. Nobody outside Anthropic can tell you exactly how Claude detects or weighs a suspicious source in every case, and anyone who claims to have a precise formula is guessing. What we can do is reason from the stated intent of the rules, from how trustworthy information has always been judged, and from the long history of search engines fighting similar tricks. That reasoning points to a fairly clear set of practical conclusions.

Why Fake Sources Became a Target

To understand why AI developers care, picture how an assistant answers a local question. It may draw on what it learned during training, and in many products it can also search the web in real time and read pages before responding. Either way, the answer is only as good as the material it reads. If someone floods the web with pages that all say "Company X is the top-rated roofer in Wake County," and those pages look independent, a naive system might conclude that many independent voices agree. In reality, one party wrote all of them.

This is not a new problem. Search engines have dealt with link farms, doorway pages, and fake review sites for decades. What is different with AI answers is the format of the output. A search results page shows ten links and leaves the judgment to you. A chatbot often gives one confident answer in a friendly voice. When that answer is wrong because it was manipulated, the person asking has fewer cues that something is off. That raises the stakes for the AI company, because the assistant's credibility is the product. An assistant that can be steered by whoever publishes the most self-serving pages becomes useless, and eventually untrusted.

There is also a safety angle that goes beyond marketing. The same techniques that might push a mediocre contractor up a recommendation list could push misleading health claims, scams, or political misinformation. Rules that make an assistant more skeptical of manufactured sources protect users across every topic. Local marketing is just one small corner of a much larger effort, but it is a corner where some vendors have been aggressive.

Finally, there is the matter of hidden instructions. Some pages have experimented with text aimed directly at AI systems, sometimes invisible to human visitors, telling the model to describe a business favorably or to ignore competing information. This kind of content is widely viewed as an attempt to hijack the assistant rather than inform it. It is reasonable to expect AI developers to treat it as a strong negative signal, and it is the kind of thing any honest business should avoid entirely.

What Counts as a "Fake Source" in Local Marketing

Many owners hear the phrase "fake sources" and assume it only applies to obvious spam. In practice, some common local marketing tactics sit uncomfortably close to the line. Here are patterns worth recognizing, whether you created them yourself or a vendor did it on your behalf.

Self-written "best of" lists. A page titled something like "Top 7 HVAC Companies in Raleigh" that lives on a site you or your agency controls, ranks you first, and never discloses the relationship. These lists are designed to look like independent journalism.
Satellite site networks. A cluster of thin websites with names like "raleigh-roof-guide" or "triangle-pest-tips" that exist mainly to link back to your main site and mention your company as the obvious choice.
Fabricated reviews and testimonials. Reviews written by staff, friends paid to post, or invented customers. These were already a problem with platforms and regulators; they are also exactly the kind of signal an AI assistant should learn to discount.
Mass-produced articles with no real expertise. Hundreds of auto-generated pages repeating the same claims with the city name swapped out, published to create the appearance of authority.
Planted forum and Q&A posts. Accounts that pose as homeowners asking "has anyone used Company X?" and then answer their own question with glowing praise.
Hidden machine-directed text. Content placed on a page specifically to instruct AI systems, especially when hidden from human readers.
Fake press releases and "awards." Paid awards from organizations that exist only to sell badges, or press releases framed as news coverage when no outlet independently covered anything.

The common thread is deception about who is speaking. A company saying "we are a family-owned plumbing business serving Cary since we opened our doors" on its own site is fine; everyone knows the company wrote it. The problem arises when content disguises its origin to borrow credibility it has not earned. If a reasonable homeowner would feel misled upon learning who actually wrote a page, that page is the kind of thing these rules are pointing at.

What This Means for Local Service Businesses

For most honest local businesses, this change is good news. If you have been doing the fundamentals well, an AI system that gets better at filtering manipulation should, in principle, give your genuine signals more room to matter. The companies that should worry are the ones whose visibility depends on a scaffolding of manufactured content.

That said, there are a few practical implications worth thinking through.

Shortcuts carry more risk than they used to

Some vendors have been selling "AI optimization" packages built on volume: publish a large number of pages and placements, get your name mentioned in as many places as possible, and hope the assistants notice. When AI systems are explicitly instructed to be skeptical of content built to sway them, that strategy looks less like an investment and more like a liability. At best, the effort is wasted. At worst, a pattern of manipulative content could make your legitimate signals look suspicious by association. We cannot tell you exactly how any particular system responds, but the direction of travel is not in favor of shortcuts.

Your reputation off your own site matters more

If AI assistants increasingly favor independent, verifiable information, then what other people say about you, in places you do not control, carries real weight. That includes genuine customer reviews on established platforms, mentions from local organizations, coverage by local news when you do something newsworthy, and listings with accurate, consistent business information. None of these can be manufactured honestly at scale, which is exactly why they are valuable.

Your own website still matters, as a clear first-party source

Your website is where you get to explain, in your own words, what you do, where you do it, and how you work. AI systems and human visitors alike benefit from a site that is specific and accurate: service areas, the types of jobs you handle, licensing information, how scheduling works, what a homeowner should expect during a visit. The goal is not to "trick" an assistant into praising you. The goal is to be the clearest, most reliable answer to the factual questions people ask about your business.

Consistency becomes a trust signal

When your business name, address, phone number, hours, and service list match across your site, your business profiles, and directories, any system trying to understand your business has an easier job. When details conflict, the system has to guess. Inconsistency is not deception, but it creates confusion, and confusion rarely helps you get recommended.

Build Signals AI Can Trust: A Practical Checklist

Here is a practical, owner-friendly list of things that build durable visibility, whether the person looking for you uses a search engine, an AI assistant, or a neighbor's recommendation. None of these depend on tricking anyone, which is precisely why they hold up when the rules change.

Ask every satisfied customer for a review, and make it easy. Send a short text or email after the job with a direct link. Do not script the review, offer payment for it, or filter who gets asked based on whether you expect praise. Real reviews in real customers' words are hard to fake and easy to trust.
Respond to reviews, including the critical ones. A calm, specific, helpful reply to a negative review often says more about your company than a dozen five-star ratings. It also shows any reader, human or machine, that a real business with real accountability is behind the name.
Write service pages that answer actual questions. What does a termite inspection involve? How long does a water heater replacement usually take? What should a homeowner do while waiting for an emergency plumber? Answer these honestly, in plain language, on your own site.
Be specific about where you work. List the towns and neighborhoods you genuinely serve. Do not create a separate thin page for every town within a hundred miles if you would never drive there. Accuracy beats volume.
Show real credentials. License numbers, insurance information, manufacturer certifications, and trade association memberships you actually hold are verifiable facts. Put them where people can find them.
Keep your business information consistent everywhere. Audit your listings once or twice a year. Fix outdated phone numbers, old addresses, and stale hours.
Earn local mentions the old-fashioned way. Sponsor a youth team, help with a community event, offer expert comments to local reporters on seasonal topics like storm damage or mosquito season. Mentions that come from real involvement are the opposite of fake sources.
Publish useful content under a real name. If your lead technician has twenty years of experience, let them explain common problems in their own voice. Real expertise, attributed to a real person, is a strong signal of trust.
Disclose relationships. If you publish a comparison or recommend a product you sell, say so plainly. Disclosure turns a potentially misleading page into an honest one.
Make your brand memorable to humans. The most durable visibility is when people already know your name before they ask anyone, human or AI. A recognizable sound, like a short custom jingle, and a consistent look help people remember and search for you directly.

Notice that almost nothing on that list is new. These are the same fundamentals that have served good local businesses for a long time. The change in AI rules mostly removes the temptation to skip them.

What to Stop Doing Now

If you are not sure what your current marketing includes, now is a good time to ask. Many owners hand marketing to a vendor and never see the details. That is understandable when you are busy running crews, managing trucks, and answering emergency calls. But you are the one whose name is on the work, and you should know what is being published on your behalf.

Here are questions worth asking your current provider, or yourself if you handle marketing in-house:

Do we own or control any websites besides our main site? If so, what are they for, and do they disclose that they belong to us?
Has anyone written reviews, testimonials, or forum posts on our behalf that were not written by actual customers?
Are we listed on any "best of" pages that we paid for or that our agency created?
Has any content been added to our site that is hidden from visitors or written specifically to instruct AI tools?
How many pages have been published on our site in the past year, and could a human expert on our team stand behind each one?
Have we bought any awards, badges, or "featured in" logos from organizations that do not actually review businesses?

If any of the answers make you uncomfortable, the fix is usually straightforward. Remove or clearly disclose self-created ranking pages. Shut down satellite sites that exist only to boost your main site. Stop any practice that puts words in customers' mouths. Delete hidden machine-directed text. Consolidate thin, repetitive pages into fewer, better ones. None of this is glamorous, but cleaning up questionable content is a reasonable step toward being on the right side of where AI systems, search engines, and consumer protection expectations are all heading.

It is also worth being skeptical of new pitches. If someone offers to "get you into ChatGPT answers" or "guarantee AI recommendations," ask exactly how. If the method involves creating content that pretends to be independent, you now have a clear reason to decline. A legitimate provider should be able to explain their approach in terms of real customers, real reviews, real expertise, and accurate information.

How Digital Marketing Raleigh Approaches This

Digital Marketing Raleigh is a done-for-you local marketing service operated by Petronella Technology Group, Inc. here in Raleigh. Our clients are mostly service businesses: pest control, HVAC, plumbing, roofing, and similar trades where the phone ringing is what matters. We built our approach around the idea that the marketing should be something an owner could read line by line and feel comfortable defending to a customer.

In practice, that means we focus on work that tends to compound over time rather than tactics that rely on fooling a system. We help clients set up simple, repeatable review requests so that genuine feedback comes in steadily. We write and maintain service pages that answer the questions homeowners actually ask. We keep business information consistent across the places people look. We help owners build a recognizable brand, including memorable audio through our jingles program, so more people search for the company by name rather than by category. And we explain what we are doing in plain language, so the owner always knows what is published under their name.

We also put our commitments in writing. You can read the terms of our guarantee to see how we think about accountability. We would rather earn trust slowly and keep it than chase a short-term spike that could disappear the next time a platform updates its rules.

Because Petronella Technology Group, Inc. has a background in technology and security, we tend to look at AI changes like this one with a practical, risk-focused eye. When an AI developer publicly states that it is tightening rules against manipulation, the sensible response is not to look for the next loophole. It is to make sure your business has nothing to hide and plenty of genuine evidence of good work. For more on how AI and search changes affect local businesses, browse our Field Notes blog.

The Bigger Picture: Trust Is the Long Game

It helps to step back and look at the pattern. Every major channel that people use to find businesses has gone through the same cycle. A new channel appears. Early adopters find ways to game it. The platform responds by rewarding genuine signals and discounting manipulation. Businesses that built on fundamentals keep their position, and businesses that built on tricks have to start over. Directories, search engines, review platforms, and social media have all followed some version of this arc.

AI assistants appear to be moving through that cycle faster than earlier channels did. The reported change to Claude's rules is one visible step, and it is reasonable to expect other AI developers to pursue similar goals, because none of them benefit from being easy to manipulate. For a local business owner, the lesson is not to panic or to chase every headline. It is to recognize that the qualities that make you trustworthy to a homeowner, such as honesty, consistency, real expertise, and a track record of satisfied customers, are increasingly the same qualities that make you visible to machines.

That is a hopeful conclusion for anyone who does good work. The homeowner with a flooded basement at midnight wants a plumber who will actually show up and fix the problem. The AI assistant helping that homeowner, if it is working as its makers intend, wants the same thing. Your job is to make sure the evidence of your good work is real, visible, and easy to verify.

Frequently Asked Questions

Will Claude's new rules hurt my business if I have never used shady tactics?

Probably not, and they may help. Rules aimed at filtering manipulative sources are designed to reduce noise, which in principle gives genuine signals like real reviews and accurate business information more room to matter. We cannot promise any specific outcome in any AI system, but honest businesses are not the target of these changes.

Is my own website considered a "fake source" because I wrote it?

No. A company describing its own services on its own website is a first-party source, and readers understand that. The concern is with content that disguises who wrote it, such as a ranking page that appears independent but was written by the business it ranks. Keep your own site accurate and specific, and you are on solid ground.

How would I know if my marketing vendor created fake sources for me?

Ask directly. Request a list of every website, listing, and page they have created or placed on your behalf. Look for satellite sites, self-written "best of" lists, reviews you do not recognize, and any hidden content on your site. A trustworthy vendor will answer these questions openly.

Should I delete old satellite sites or ranking pages?

In most cases, yes, or at minimum disclose clearly that you own them. Content that exists mainly to make your business look independently endorsed works against the direction AI systems and search engines are heading. Many owners find that consolidating into one strong, honest website is simpler to maintain anyway.

Can anyone guarantee that my business will appear in AI answers?

Be cautious of anyone who promises that. AI systems decide what to say based on many factors that outside parties cannot control or fully see. A provider can help you build the genuine signals that tend to support visibility, but a guarantee of placement inside a specific AI product is not something anyone can honestly make.

Do reviews still matter if AI tools are summarizing information?

Genuine reviews arguably matter more. They are independent, written by real customers, and difficult to fake at scale without detection. When an assistant tries to understand whether a local business is reliable, real customer experiences are exactly the kind of evidence a careful system should weigh.

What is the single most useful step I can take this month?

Set up a simple, consistent way to ask every satisfied customer for a review, and commit to replying to every review you receive. It costs little, it builds genuine trust signals, and it keeps paying off no matter how search engines and AI assistants change their rules.

If you want help reviewing what is currently published under your business name, or you want a local marketing partner that builds visibility on genuine signals instead of shortcuts, call Digital Marketing Raleigh at 919-348-4912 or reach us through the contact page. We will walk through where you stand and what makes sense for your business.