Why customize your assistant#

Using ChatGPT or Claude in generalist mode is like hiring a brilliant intern who knows nothing about your brokerage, your market, your clients, or your way of working. They can do a lot, but each task requires starting explanations from scratch.

A custom AI assistant is the opposite. It’s a teammate who has absorbed your templates, your tone, your typical listings, your market data, your responses to classic objections, and reuses them automatically in every interaction. For the bigger picture, see generative AI applied to real estate.

Two main options in 2026:

1. A custom GPT via ChatGPT (OpenAI): perfect for public use and easy to share
2. A Claude Project via Claude (Anthropic): superior for document analysis and advanced professional uses

Method 1: build a custom GPT#

Photo: MART PRODUCTION · Pexels

Custom GPTs are configured versions of ChatGPT you create for a specific use. Available with a ChatGPT Plus subscription (~$20/month). You can keep it private, share with your team, or make it public for your buyers.

Step 1: access the creator#

Sign in to chat.openai.com with a Plus account. Top left, “Explore GPTs,” then “Create.” The interface is guided and intuitive.

Step 2: name and role#

Give it a clear name: “[Your brokerage name] Assistant”. Description (1 line): “AI assistant for [brokerage], specialized in listing writing, client emails, and real estate document analysis.”

Step 3: write the instructions (the GPT’s heart)#

This step determines 80% of the quality. Here’s a template to copy and adapt.

You are the AI assistant for [Brokerage Name], a real estate brokerage
specializing in [market/property type] in [city].

PRIMARY ROLE
You help the brokerage agents to:
- Write punchy, personalized real estate listings
- Compose professional emails
- Analyze documents (HOA minutes, leases, energy disclosures, purchase agreements)
- Prepare pitches for seller appointments
- Create content for social media

IDENTITY AND TONE
- Positioning: [boutique brokerage / high-end / investment expert]
- Tone: [warm and professional / expert and factual / modern and accessible]
- Avoid: hollow jargon, unverifiable promises, overly salesy phrasing

WHAT YOU DON'T DO
- Invent prices, square footage, or market data
- Generate contracts with legal value
- Promise guaranteed timelines or results

If information is missing to answer well, ask for it clearly
rather than improvising.

Step 4: upload the knowledge base#

In the “Knowledge base” tab, upload:

• The brokerage book (sales presentation, PDF or Word)
• Your email templates
• A market data document: price per sq ft by area, sale times, trends
• Your responses to the 10 most frequent objections

ChatGPT accepts up to 20 files of 512 MB, more than enough.

Step 5: test and refine#

Before sharing, test on 5 to 10 real cases:

• “Write a listing for a 770 sq ft 2-bed on Lafayette Street SoHo, courtyard view, 3rd floor walk-up, target young couple”
• “Follow-up email for Mr. Smith who toured Park Avenue yesterday and is hesitating on budget”
• “3 arguments to convince a seller their $650K price is 10% overvalued”

If responses are too generic, enrich the instructions. If the tone doesn’t fit, adjust the “Identity and tone” section. Count 2-3 refinement sessions before a truly calibrated GPT.

Step 6: share#

Via the share interface, give access to your GPT to all brokerage members (ChatGPT Plus account). It’s the simplest way to standardize communication.

Method 2: build a Claude Project#

Claude Projects offer a powerful alternative with a marked superiority on document analysis. The complete setup is detailed in our guide to configure your Claude Projects. In summary:

1. Create a project on claude.ai (Pro subscription, ~$20/month)
2. Add your permanent instructions
3. Upload your reference documents into the project base
4. Access this assistant at each new conversation, Claude has the context automatically

When to prefer Claude Projects#

For analyzing your HOA documents, long leases, complex disclosures, multi-document files. Claude is more rigorous and less likely to hallucinate on technical documents.

When to prefer a custom GPT#

For daily writing tasks, emails, social media posts. GPT-4o is slightly more creative and fluid on real estate storytelling.

In practice, the most advanced brokerages use both: GPT for creative writing, Claude for document analysis.

Advanced use case: the 24/7 buyer assistant#

Photo: Yan Krukau · Pexels

One of the most powerful applications: build a GPT for your buyers, not just your team.

Concretely, you configure a GPT that knows your portfolio listings (up-to-date list, re-uploaded regularly), your area’s characteristics, and the answers to frequent buyer questions. You share access with your qualified buyer network.

Your buyers can ask questions outside business hours, “Does this property have parking?”, “What’s the living room square footage?”, “Are there any HOA assessments voted?”, and get an immediate response based on your real data. You’re notified of conversations requiring your direct intervention.

Commercial impact: a buyer who gets a precise response at 10:30 PM perceives you as a modern, responsive brokerage. You no longer lose leads because you weren’t available at the right moment.

The assistant that knows your brokerage is the agent who never sleeps. It doesn’t sign listings for you, but it keeps the conversation open when you aren’t there.

The complete instructions template#

Here’s the extended version, to use as a starting point for your GPT or Claude Project.

IDENTITY
You are the AI assistant for [Name], a real estate brokerage in [city/area].
Founded in [year], the brokerage specializes in [specialty].

MISSION
Help agents to: write listings, emails, social posts, seller pitches, document analyses.
Help buyers to: find info on portfolio listings, understand the buying process (if public GPT).

COMMUNICATION RULES
- Always use polite/professional address with external clients
- Use casual address with brokerage agents
- Never promise a precise price or timeline without data
- Rephrase vague questions
- Always propose a concrete next step

WRITING STANDARDS
- Listings: 180-220 words, no hollow adjectives, precise target angle, strong hook
- Emails: max 200 words, short direct subject line, 1 clear CTA at the end
- Instagram posts: 80-100 words + 1 question, 5-8 relevant hashtags

LOCAL MARKET DATA
[Summary of your key data: price per square foot by area, sale times, trends]

PORTFOLIO LISTINGS
[If applicable: list of current listings with brief description]

WHAT YOU NEVER SAY
- "I can't help with that" → always propose an alternative
- Mention competitors by name
- Invent factual data if you don't have it

Adapt each bracket to your brokerage’s specifics. The “Local market data” and “Portfolio listings” sections need monthly updates—that’s what distinguishes an average assistant from an excellent one over time.

The mistakes to avoid#

• Instructions too short: under 200 words, the assistant doesn’t have enough context. Aim for 500-800 words
• No updates: a GPT loaded once and forgotten becomes obsolete. Reload files monthly (listings, market data)
• Mixing too many tasks: a GPT for listings + a GPT for document analysis work better than one generic GPT that does everything
• No real testing: a GPT tested only by the broker doesn’t reflect agent usage. Have 2-3 different people test before sharing

The next step#

Once your assistant is configured and tested, the next step is integrating it into the whole team’s workflow. Our program to train your team on AI in 4 weeks covers exactly this, with concrete deliverables and a week-by-week schedule.

The custom AI assistant isn’t a technical project reserved for startups. With current ChatGPT and Claude interfaces, it’s accessible in 10-30 minutes of initial configuration. The investment pays off immediately: an AI assistant that knows your brokerage produces usable results in seconds, where generalist AI needs 3 exchanges to understand context.

In 2026, the difference between a modern brokerage and a lagging brokerage often comes down to these small systematic advantages: minutes saved per interaction, more consistent content, increased responsiveness. The custom AI assistant is one of the simplest levers to put in place to get there.