Anthropic's Deceptive Campaigns Rule: What It Means for Agent Social

Anthropic's updated Usage Policy, which took effect on November 12, rules out using Claude to run fake accounts, write fake reviews, or post content that hides who is really behind it. For real estate social media, that means no made-up commenters on listings, no AI-written testimonials presented as a client's words, and no invented agent bylines.

Most agents will never come near that line. Plenty drift toward it without noticing, because the shortcuts sound ordinary: a second account to give a new listing early likes, a review you drafted for a happy client who said yes, a friendly made-up name on the blog. We take each one below against the policy text and Google's own guidance. Our social media guide covers the daily basics. This post covers the line.

What the new section says, in plain terms

Anthropic announced the update on October 8 in its 2026 Usage Policy update and said the updated policy takes effect on November 12. The activity was already prohibited, but the rules were scattered across four sections: elections, fraud, privacy and disinformation. They now sit in one section, Do Not Engage in Deceptive Campaigns or Artificial Activity, which applies to deceptive activity of any kind, political or commercial.

The Usage Policy itself lists six things under that heading. The ones an agent can bump into:

• Running fake personas or accounts to mislead people about where something came from, how many people hold a view, or if they are talking to a human. Named examples: sockpuppets, astroturfing, fake reviews and bots that claim to be human.

• Hiding who sponsors content meant to influence public opinion or decision-makers.

• Spreading content through networks of seemingly independent sites or accounts to hide a common source.

• Seeding search engines or AI systems with content that misrepresents its origin, authorship or independence.

• Building tools for these campaigns, or selling them as a service. If a vendor sells you comments, followers or reviews, ask where they come from.

One honest limit: this section is about deception, not about using AI to write. The policy also governs only Anthropic's products. Instagram, Facebook and Google have their own rules, and consumer protection law is a question for your broker or counsel. What follows is our reading, not legal advice.

Myth 1: "A second account to boost our listings is harmless"

The myth. You open a new Instagram account, have it like and comment on the new listing within the first hour, and call it a head start.

The fact. The word that matters is independent. A second account that is openly yours is ordinary marketing: a listings account or a team account with the brokerage name in the bio. The policy targets accounts that pose as someone else to amplify content. A "neighbor" who comments "just toured this one, wow" on a house nobody toured is a sockpuppet, and the policy uses that word.

In practice: an assistant has Claude write forty varied comments for a new listing and posts them from three accounts with invented names. That is the pattern the policy describes. The version that works: the agent, the team and the seller's real friends share the post from real accounts.

Myth 2: "AI-written reviews are fine if the client agrees"

The myth. The client loved the closing, said "sure, write whatever," and you paste a polished paragraph into Google under their name.

The fact. A review is only worth something because the client wrote it. Fake reviews are named in the policy's first bullet. Anthropic's page does not discuss the client who agrees. Our reading: when the words are not the client's own, the reader is misled about who is speaking, and permission does not change what the reader sees.

What AI can do honestly is help with the part that is yours:

1. Draft the request message you send after closing, in your voice.

2. Give the client three short prompts (what worried you, what surprised you, who would you tell) so they write their own lines.

3. Draft your public reply to a review you received, posted under your name.

The client types the review. You write the thank-you. Our post on the never-say list and the disclosure line covers wording rules for the posts themselves.

Myth 3: "A made-up agent name and AI headshot is fine for blog posts"

The myth. The blog needs a byline, nobody has time, so posts go out under a friendly invented name with a generated face.

The fact. Google says it strongly encourages "accurate authorship information, such as bylines to content where readers might expect it," in its guidance on helpful, people-first content. The same page names fabricated creator profiles as deception, with AI-generated headshots, made-up names and false credentials as examples, and says any form of deception makes a page untrustworthy to users and to Google's quality systems. Anthropic's first bullet covers the same ground: personas built to mislead people about origin. A fake byline is a persona on a page.

Google also frames content around who made it, how, and why. The honest answers are easy: the agent named on the post read it before it went live, AI help is noted where readers would reasonably expect it, and the post exists to help a buyer or seller decide something.

Myth 4: "Five local sites, each with its own voice, spread our reach"

The myth. Neighborhood pages on five separate sites, written by AI in five different "local voices," each linking back to your listings page.

The fact. This is the third and fourth bullets in one move: content spread through seemingly independent sites to hide a common source, and material placed where search engines and AI assistants draw answers. Google's spam policies add their own warning. Scaled content abuse, as Google defines it, is many pages generated for the primary purpose of manipulating search rankings and not helping users, no matter how the content is created.

Google does not say AI-written pages are the problem. Volume aimed at rankings is. One site, a named author, and a page that answers a real question holds up under both sets of rules.

Myth 5: "Our chat bot should sound like one of the team"

The myth. An AI assistant answers Instagram messages and website chat using a team member's first name and photo so it feels personal.

The fact. The first bullet includes bots that claim to be human. An assistant can answer "is the open house still on Sunday" at 10 PM. It cannot pretend to be a person to do it. The version that works:

• Name it plainly, such as "Assistant for [your brokerage]," and say in the first message that it is an AI.

• Limit it to facts you have approved: event times, how to book a showing, where to find a listing.

• Hand off to a real agent for anything about price, negotiation or advice.

By the numbers

Every figure behind this post:

• November 12: the date the updated Usage Policy takes effect, per Anthropic's announcement, which was published October 8.

• Four: the older policy sections whose scattered rules moved into the new section.

• Six: the items listed under Deceptive Campaigns in the Usage Policy.

A five-minute audit for your accounts

1. List every account, site and profile that posts for your brokerage or any agent. Write the real person responsible next to each.

2. Mark any account whose bio does not say who runs it.

3. Check every blog byline. Each should be a real person who read the post before it went live.

4. Ask each vendor in writing where their followers, comments or reviews come from.

How Urban Marketing Edge handles this

We use AI to move faster on drafts and variations. Strategy and taste stay human, and every post has a real name behind it. Our post approval system means nothing goes out unseen, so a broker or team lead can catch anything that crosses these lines before it publishes. For Rocks Realty and Clareo Group, the work is real content from real agents.

We do not sell followers, comments or reviews, and we do not run accounts that pretend to be independent. Our social media management service shows how that works from drafting to approval.

Questions brokers ask

Does Anthropic's new usage policy apply to my brokerage?

It applies to anyone who submits inputs to Anthropic's products, including people in the Claude app, developers on the API, and end users of products that integrate Claude. If your team uses Claude directly or through a tool built on it, the policy covers that use.

Can agents still use AI to write social media captions?

Yes, by our reading. The Deceptive Campaigns section is about fake accounts, hidden sponsors and misleading origin, not drafting. A caption an agent reviews and posts under their own name does none of those things.

Is a second Instagram account allowed?

An account that is openly yours, with the brokerage or team named in the bio, is ordinary marketing. An account that poses as an independent person to like, comment or share your posts is the pattern the policy describes.

Can I use AI to write a reply to a client's review?

Yes. A reply is your message, posted under your name. What the policy rules out is writing the review itself as though the client said it. Read it, then sign it as yourself.

Do I need to tell readers a post was written with AI?

Google suggests disclosure where readers would reasonably expect it, and asks if AI use is self-evident. Make one office rule and apply it to every agent.

Book a 30-minute strategy call

If your audit turned up an account nobody can explain, a vendor you cannot vouch for, or blog posts with no real author, bring the list to a call. We will map the simplest way to keep content moving with a real name on every post and a clear approval step.

Book a 30-minute strategy call and we will start with your accounts. You can also review our plans to see how monthly social, approvals and reporting fit together.