Prompt Expansion for Real Estate: Why Most Broker AI Output Is Generic, and How to Fix It

There's a quiet split happening between brokers who get useful output from ChatGPT and brokers who don't. It isn't about which model they use. It isn't about whether they pay for the Pro tier. It's about how they ask.

Brokers in the second group treat ChatGPT like a search engine. They type "find me F&B tenants expanding in SoHo" and get back a list of generic chains they already know about. Brokers in the first group treat ChatGPT like a junior analyst. They tell it the role, the market, the deal size, the format they want, and the constraint that prevents fluff. They get back a ranked list with reasoning that's actually useful.

That's the gap a prompt expander closes.

Prompt expansion takes a short, conversational request from a real estate broker and rewrites it as a complete prompt for ChatGPT, Claude, or Perplexity. A good expanded CRE prompt establishes the broker's role and market, states the task in concrete terms, names the asset class and geography, requests a specific output format, and adds constraints that prevent generic AI fluff. Station CRM's free prompt expander tool is purpose-built for commercial real estate workflows: prospecting, market research, outreach, due diligence, OM writing, and lease review. The broker types the way they'd ask a colleague. The tool fills in the structural prompt engineering automatically. Output comes back in under five seconds, ready to paste into any LLM.

What a Good CRE Prompt Looks Like

Pick any AI prompting guide and it'll tell you the same five things. State the role. Provide context. Be specific about the task. Specify the output format. Add constraints.

Now look at how most brokers actually prompt:

"list landlords in SoHo I should call about a tenant"

That's a request, not a prompt. The AI has no idea what kind of tenant, what size, what budget, what the broker's hook is, or whether the output should be five names or fifty. It produces a generic answer and the broker decides AI isn't useful for prospecting.

Compare to the expanded version:

"You are a NYC retail leasing broker covering SoHo. I have a fast-casual F&B tenant looking for 1,500 to 2,500 SF, ground floor, willing to pay up to $300/SF. List 12 SoHo landlords whose portfolios match this profile and who have had vacancies or recent tenant turnover in the last 18 months. For each, include: portfolio properties in SoHo, recent tenant activity, who handles leasing decisions, and one specific reason this tenant is a fit. Format as a ranked list. Skip any landlord you can only confirm from generic press releases."

The second prompt produces something a broker can act on. The first one wastes ChatGPT's time and the broker's.

Why Generic Prompt Tools Fall Short for CRE

There are dozens of prompt expansion tools online. Most are built for marketing teams, copywriters, or general business use. They expand "write a blog post about X" into a structured marketing brief. That's useful for a content team. It's not useful for a tenant rep trying to identify expansion candidates in Brooklyn.

CRE prompts have specific conventions:

They have to specify market and submarket. "NYC retail" is not the same as "Manhattan retail" and neither is the same as "Bedford Avenue corridor in Williamsburg." A useful broker prompt names the geography precisely.
They have to name the asset class. The expansion thesis for a fast-casual restaurant is different from the expansion thesis for a flagship apparel brand or a fitness operator. ChatGPT will conflate these unless you don't let it.
They have to specify deal size. A tenant looking for 800 SF is not the same as a tenant looking for 8,000 SF, and the relevant landlords are different.
They have to identify the broker's side. Tenant rep prompts and landlord rep prompts produce opposite output. So do investment sales and leasing prompts.
They have to constrain output format. Without that, AI tools default to long, hedged, unhelpful prose.

A general-purpose prompt expander doesn't know any of this. A CRE-specific one does.

The Seven Common Broker Tasks That Benefit Most

After watching brokers actually use AI tools, the same seven categories of work come up over and over. The Station CRM prompt expander is built around them:

Prospecting. Identify and qualify landlords, tenants, brands, and owners. Good expanded prompts ask the AI to surface specific signals (closings, expansion announcements, ownership changes, 1031 windows) and rank the output with reasoning. This is the most useful AI category for most brokers because it turns a research task that would take an hour into something that takes five minutes.

Market research. Build a market or submarket picture with asking rents, comps, vacancy, tenant mix, and recent deals. Good expanded prompts request sourced figures, time windows, geographic boundaries, and what's changing. Output comes back as a brief with sections, not a wall of text.

Outreach emails. Draft cold and warm broker emails. Good expanded prompts specify recipient role, the broker's hook, urgency, call to action, and word constraint. The default ChatGPT email opens with "I hope this email finds you well." The expanded version doesn't.

Due diligence. Pull together what a broker needs to know before pursuing or recommending a property or tenant. Good expanded prompts request ownership history, zoning, lease comps, tenant credit, neighborhood trajectory, and red flags as a checklist.

OM and property marketing copy. Draft offering memorandum prose or property descriptions. Good expanded prompts specify the audience (institutional buyer, owner-user, tenant rep), the property type, the most important attributes, and tone.

Lease and LOI review. Read a lease or term sheet and surface what matters. Good expanded prompts ask the AI to identify economic terms, escalations, obligations, options, and unusual provisions, then output a structured summary with flagged issues.

Anything else CRE. The catch-all. Meeting prep, comp analysis, broker BOV drafting, market reports, the long tail of one-off requests.

Which AI Tools the Expander Works With

Short answer: all of them. The expander produces plain-text structured prompts, the kind every modern LLM is tuned to handle. There's no special formatting, no embedded markup, nothing model-specific. Paste and go.

The list of tools brokers are actually using in 2026 has gotten long. Here's the working set:

General chat models, the ones most brokers paste into

ChatGPT (GPT-4, GPT-4o, GPT-5, the o-series reasoning models)
Claude (Claude.ai, Claude Pro, Claude Teams)
Claude Cowork (Anthropic's collaborative workspace, more brokers are landing here through their firms)
Gemini (Pro, Ultra, Advanced)
Grok (xAI, useful for X/Twitter-flavored research)
Microsoft Copilot (M365, Word, Outlook, Teams, Bing Chat)
Perplexity (Sonar, Pro, Spaces, the research-leaning option)
Mistral / Le Chat
DeepSeek (V3, R1, especially if cost matters)
Pi (Inflection, the conversational option)

Open-weight and emerging models

Llama (Meta 3.1, 3.3, 4)
Hermes (Nous Research, Hermes 3 and 4)
OpenClaw
Cohere Command
Reka
Qwen (Alibaba)
Manus (the Chinese agent framework that's been catching on for research-heavy tasks)

Embedded AI in tools brokers already use

Notion AI (for OM and brief drafting)
Gmail / Outlook AI compose (for outreach drafts)
HubSpot AI breeze
Salesforce Einstein

What about your firm's internal LLM gateway? Yes, those work too. If your brokerage runs Claude or GPT through a private API gateway with company prompts and policies, the expanded prompt you paste in still respects those. The expander just gives you the structural starting point.

The reason all of these work is that good prompt engineering is structural, not model-specific. Role, context, asset class, output format, constraints, that's the recipe. ChatGPT, Claude, Manus, OpenClaw, Hermes, all of them respond better when those pieces are present. The expander does that structural work; the underlying model handles the reasoning.

Where This Sits Relative to Other Station CRM Tools

The prompt expander is one of three AI prompt tools Station CRM publishes for free.

The AI prompt library is 32 hand-written, fill-in-the-blank prompts for specific CRE workflows. Pick a template, fill in the blanks, paste into ChatGPT. Useful when the situation matches a template.

The Claude Skills library is 15 system prompts that turn Claude into a persistent CRE assistant. Set it once and Claude behaves like a market analyst, LOI drafter, or deal coach across an entire conversation.

The prompt expander is for everything that doesn't fit a template, which is most of what brokers actually do. You describe the situation in your own words. The expander builds the prompt.

For brokers who want the structural prompt engineering done automatically without losing their own voice, the expander is the path of least resistance.

What Prompt Expansion Doesn't Do

It doesn't do the AI work. It builds the prompt; the broker still pastes it into ChatGPT, Claude, or Perplexity and reviews the output. The expander is not a replacement for AI judgment, it's a layer that makes the AI tool produce broker-grade output instead of generic content.

It also doesn't replace what a real CRM does. A useful CRE prospecting prompt depends on knowing things about your pipeline, your relationships, and your market that ChatGPT will never know. That's where Station CRM's AI chief of staff sits, in the layer below: the system that knows your deals, your contacts, the closings happening this week, and the 1031 sellers in their identification window. The prompt expander helps you get more out of generic AI tools. The AI chief of staff is the thing that knows your business.

Most brokers benefit from both. Use the prompt expander for one-off tasks where you'd otherwise type into ChatGPT. Use the chief of staff for the daily intelligence layer that ties to your actual pipeline.

Try It

The prompt expander is free and there's no signup. Type a short request, pick the task type, set the market and asset class, get the expanded prompt back, paste it into your AI tool of choice. Five seconds, no friction.

If you find yourself using it daily, the deeper thing to look at is what an AI-native CRM looks like for the work you do every week. Request a Station CRM demo to see how the prospecting, market intelligence, and chief of staff pieces fit together.