Real estate is a local business. Not locally adjacent, not regional — genuinely, granularly local. The difference between a listing in one neighborhood and a listing three blocks over isn't just geography; it's school zones, commute patterns, flood elevation, street character, and fifteen years of price history that every experienced agent carries in their head and almost nowhere in writing.
Claude carries none of it. Not because it can't — because nobody told it.
The output that sounds like everywhere
Ask most agents what happens when they use Claude for listing copy and you'll hear a version of the same story. The output is grammatically fine. It hits the obvious adjectives. "Charming," "open concept," "natural light," "minutes from downtown." It reads like a listing. It could be a listing anywhere in the country, for any brokerage, written by anyone.
That's the problem. The thing that differentiates a brokerage — its voice, its market knowledge, its read on the specific buyer for this specific property — doesn't make it into the output, because it wasn't in the input. Claude defaults to the median listing description it was trained on, which is approximately every listing description ever written, averaged together.
This is fixable. But it requires something most agents don't do: building the context that makes generic output impossible.
What real estate language work actually covers
Listing copy gets the most attention, but it's a fraction of the language work a real estate professional produces in a given month:
- CMA narratives. The written interpretation of a comparative market analysis — what the numbers mean, what the market is doing, why this price makes sense.
- Client outreach. Prospecting letters, follow-up emails, check-ins with past clients, market update newsletters.
- Offer communications. Cover letters for offers, escalation clause explanations, negotiation position summaries.
- Neighborhood and market content. Blog posts, social copy, content that demonstrates local expertise to potential clients.
- Transaction summaries. Post-closing summaries for client records, team notes, referral follow-ups.
All of it benefits from local context. All of it suffers from generic defaults. And all of it is repeatable enough that building the right context once pays dividends across every piece of output for years.
The differentiator for an independent brokerage isn't just local knowledge. It's local knowledge at scale — consistently present in every document the team produces, not just the ones the top producer touches.
The agent hoarding problem
In most brokerages, AI adoption looks like this: one or two agents figure out a workflow that works reasonably well for them. They've refined their prompts over several months. They get better output than everyone else. They might share a tip occasionally, but the knowledge is personal — it lives in their head, in their chat history, in a notes document no one else sees.
The brokerage's AI capability is therefore a function of individual initiative, not institutional knowledge. The top producer has a system. Everyone else is starting from scratch. New agents joining the team get no benefit from what the experienced agents have figured out.
How most brokerages run AI
- Each agent develops their own prompt habits in isolation
- Output doesn't reflect brokerage voice or market expertise
- New agents start from zero
- Listing copy reads generic; CMA narratives are boilerplate
- Market knowledge exists in heads, not in the system
What consistent brokerage AI looks like
- Market context and brokerage voice loaded before every session
- Listing copy reflects the actual neighborhood, buyer profile, and price point
- New agents get the same output quality as veterans from day one
- CMA narratives already know the local inventory conditions
- The brokerage's expertise is in the files, not just in senior agents
What local context actually means in practice
Real estate AI works well when the tool knows things like: this brokerage focuses on the mid-city and garden district markets; our typical buyer is a professional household, 30–45, buying in the $350k–$650k range; here's how we describe walkability vs. how a national template would; here's the school district situation and how we talk about it; here's our voice (informed but not pretentious, local but not provincial); here's what we never say because it sounds like everyone else.
None of that is secret. It's just knowledge that lives in the heads of experienced agents and nowhere else. Moving it into a form the tool can use — once, properly — changes the quality of every output the team produces from then on.
Pull three listing descriptions your top producer wrote last year. Pull three AI-generated descriptions from an agent who hasn't set up any context. If you can't immediately tell which is which based on voice, market specificity, and knowledge of the buyer — the context problem is solved. If you can tell immediately, you know what's missing.
Field notes
Real estate has a version of the same problem every service business has with AI: the valuable part is the local, specific knowledge, and the tool defaults to the generic. The fix isn't a better prompt. It's capturing the local knowledge in a form the tool can load before the session starts. Most brokerages have that knowledge sitting in the heads of their experienced agents. The work is getting it out.
R.P.
