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.

What’s Claude Code, for non-developers#

Claude Code is a command-line tool developed by Anthropic. It installs in your terminal (Terminal on Mac, PowerShell on Windows) and lets Claude execute code directly on your computer for you.

You don’t see the code. Describe your need in English, Claude:

• Download and process data files (CSV, JSON, Excel)
• Write and execute Python or JavaScript scripts
• Generate PDF, Excel, Word documents automatically
• Access public APIs (including MLS public records)
• Create dashboards with charts

The difference with web Claude: there, Claude explains how to do it. Here, Claude does it for you. For the general context of this AI tooling at the brokerage, see how AI transforms day-to-day.

Why MLS data is a goldmine#

In the US, MLS data is the gold standard, and county property records are publicly accessible. Together, they contain the history of millions of US real estate transactions: address, square footage, price, sale date, property type. Sources: your local MLS, ATTOM Data, Reonomy, Estated, Zillow Public Records, county assessor sites.

Most agents consult them occasionally. Used systematically, they let you:

• Identify owners who bought 5-10 years ago in a specific area. A 2-bed bought in 2016 in Brooklyn at $620K is worth $750-$800K today. The owner may have had kids since and not yet considered reselling. You can be the first to call them
• Detect hot micro-markets. Certain streets, certain buildings see their price per square foot rise faster than average. Identifying it before everyone else gives an unbeatable pitch at the seller appointment
• Build sectoral market reports to present at seller appointments, with data your competitors haven’t compiled

For the bigger picture on what generative AI applied to real estate changes, we’ve published a complete guide: Claude Code is just one advanced building block.

Install Claude Code, 5 minutes#

Prerequisites: a Mac (Terminal) or Windows (PowerShell), a Claude Pro or Max subscription (~$20/month), Node.js installed (download at nodejs.org, 2 minutes).

Open your terminal and type:

npm install -g @anthropic-ai/claude-code

Wait for the install to finish, then:

claude

Claude Code launches. You see a prompt. You can now give it instructions in English. That’s it. Under 5 minutes to be operational.

Case 1: extract MLS transactions for an area#

Photo: Tima Miroshnichenko · Pexels

The goal: the list of all condo transactions between 650-1,000 sq ft in a neighborhood, over 3 years, with price per square foot, address, and date.

Download MLS data for Brooklyn (zip 11215) for condo transactions between 650 and 1,000 sq ft since January 2022. Calculate the average price per square foot per half-year, identify the 10 streets with the highest price per square foot, and generate an Excel file with this data. Use the public records API (or ATTOM Data) to fetch the data.

What Claude Code does: writes a Python script, calls the public records API, filters the data per your criteria, calculates statistics, generates an Excel file on your desktop. All in under 2 minutes.

You get: an Excel table with the filtered transactions, prices per square foot by street and half-year, ready for a seller appointment. Cost of an equivalent analysis ordered from a vendor: several hundred dollars. With Claude Code: the time to type the instruction.

Case 2: generate a market report PDF#

The goal: a professional 4-6 page document you hand to the seller at the CMA appointment.

Based on the MLS data you just analyzed, generate a market report PDF with: – a title page with my brokerage name [brokerage name] – a “Local Market Analysis” section with key data and trend chart – a “Your Property in Market Context” section with price positioning based on square footage [indicate square footage] – a “Recommendation” section with a realistic price range – a professional layout The file should be named market-report-[address]–[date].pdf

You get: a PDF with real data, auto-generated charts, and an argued price positioning. The kind of document your competitors don’t present. To integrate this report into an overall Claude setup, see our guide to run your brokerage with Claude via Projects.

Case 3: build a targeted prospecting spreadsheet#

The goal: automatically identify potential sellers in an area, ranked by priority.

From Brooklyn MLS data, create a prospecting spreadsheet. Selection criteria: – condo transactions between 2014 and 2017 (7-10 years of ownership) – square footage between 600 and 870 sq ft – purchase price under $620,000 For each property: calculate estimated capital gain at current neighborhood price per sq ft. Sort by descending capital gain. Include address, purchase date, purchase price, current estimate, and estimated capital gain. Export to Excel.

You get: a qualified prospect list, sorted by capital gain potential. Owners at the top have the best statistical reasons to sell now. That’s your prospecting list of the week. Reproducible on any area, in minutes, with real and current data.

The competitive advantage isn’t AI, it’s your Monday-morning prospecting list while your competitors still work by gut feel.

Claude Code’s limits, and when to stay on the web interface#

Claude Code is powerful, but it has limits to know.

• It needs a stable connection. Scripts that download large data can fail on a slow connection
• Public APIs sometimes change. If a script fails, flag it to Claude Code: it diagnoses and fixes
• For common tasks (emails, listings, document analyses), the web interface remains faster and more intuitive
• For one-off MLS analyses, tools like Zillow Public Records or HouseCanary suffice. Claude Code becomes relevant when you want custom, repeated analyses on criteria those tools don’t offer

Who it’s really for#

Photo: Christina Morillo · Pexels

Claude Code is for you if:

• You’re comfortable with your computer (no coding needed)
• You regularly prospect and want to automate data analysis
• You present market reports at seller appointments and want to make them stronger
• You manage 30+ listings and want to monitor the market continuously

It isn’t suitable if you’re new to AI. First master the fundamentals and Claude Projects, then move to Claude Code when you’re comfortable.

The complete workflow of a modern brokerage#

Claude Code doesn’t replace the regular Claude interface, it complements it. A typical workflow:

1. Monday morning: Claude Code extracts and analyzes the area’s MLS data, generates the report and the week’s prospecting list
2. Monday afternoon: Claude (web interface, Project configured with your brokerage profile) writes targeted prospecting emails from the results
3. Tuesday to Friday: ChatGPT prepares your pitch for the week’s appointments, Claude analyzes HOA minutes of properties being toured
4. Throughout: Kappn generates immersive videos for new listings

You walk into appointments with data your competitor doesn’t have. Per Morgan Stanley, 37% of tasks at large real estate companies are automatable. Claude Code is the tool that gets you to that number in an individual brokerage, no IT department, no software budget, no outsourcing.

When Claude Code stops being optional for top brokerages#

Most US real estate agents will never need Claude Code. They will use ChatGPT, maybe Claude through the web interface, and that will cover 95% of their needs. But for the top 5% of brokerages — the ones running their own MLS analytics, doing serial cold outreach to neighborhoods, generating reports across hundreds of comparables — Claude Code crosses the threshold from “nice-to-have” to “competitive necessity.” Understanding when that threshold gets crossed lets you position your brokerage at the right moment, rather than falling behind a competitor who got there first.

The threshold is reached when three conditions stack up: you generate the same type of report more than once a week, the data sources behind that report change frequently, and the manual time it takes to build it is over an hour each time. Once those three conditions are true, Claude Code starts saving 10-15 hours per week of repetitive work, which is enough to justify the learning curve for anyone serious about scale. Below that threshold, the time spent learning Claude Code is better invested elsewhere.

What changes for brokerages that cross the threshold is structural. They stop being limited by the bandwidth of their analyst — or, if they’re a smaller brokerage, by the bandwidth of the broker themselves. They can run a comparative market analysis on every single new listing in their farm area, automatically, every morning. They can detect off-market opportunities by cross-referencing public records with MLS history, scripted to run nightly. They can produce custom seller decks tailored to each appointment in 90 seconds rather than 90 minutes. The compound effect across a year is the difference between a brokerage that produces 80 listings and one that produces 200, with the same headcount.

The second-order effect is recruiting. Top agents follow infrastructure. Once your brokerage runs on Claude Code-powered workflows, you attract the kind of agents who choose their broker based on tooling, not just commission split. That’s a long-term moat that’s difficult to replicate by a competitor who’s still doing manual MLS analysis.

Why start now, not in 6 months#

97% of US agents use AI (WAV Group, January 2026). But only 23% have rebuilt at least one core workflow around it (NAR 2026). The depth gap is the opportunity: every listing won by a brokerage that responds faster, presents better, and dedicates more time to the relationship is a listing potentially lost by an agent still working like it’s 2015.

You don’t need to be a tech expert. You don’t need to understand how AI works. You need to know what to type to get a useful result. That’s it. For the bigger picture, see generative AI applied to real estate.

What you need to start#

One tool: a ChatGPT account. Go to chat.openai.com, click “Sign up,” enter your email and create a password. Free. The free version is more than enough for the 3 use cases in this guide. You’ll see later if ChatGPT Plus (~$20/month) is worth it for faster, more precise results.

Account created? Let’s move on to the 3 use cases.

Case 1: write a tough email in 30 seconds#

The problem: you have to follow up with a seller who hasn’t responded in 3 weeks. The property is overpriced. You don’t want to seem pushy, but reopen the conversation. You reformulate 3 times, it takes 20 minutes.

The AI solution: give ChatGPT the context, it writes the email for you.

I’m a real estate agent. I visited the Smith property 3 weeks ago for a CMA. They listed their apartment at $650,000, but I think it’s worth $590,000 based on neighborhood comps. I haven’t heard from them since our appointment. Write me a professional, warm follow-up email that reopens the conversation without seeming pushy, and subtly proposes discussing the price. 120 words max.

What you get: a complete, professional email, with a subject line, in the right tone. You proofread, adjust two words, send. Total time: 45 seconds. Repeat with a real email from your week—buyer follow-up, objection response, showing confirmation. The difference is immediate.

Case 2: generate a punchy listing#

Photo: Christina Morillo · Pexels

The problem: writing a listing takes 20-45 minutes for a careful agent. The result is correct but generic: same formulas, same hollow adjectives, same structure.

The AI solution: give the property info, ChatGPT generates a structured, differentiated listing in 15 seconds.

You are an expert real estate copywriter. Write a listing for the following apartment: 730 sq ft 2-bed, 3rd floor with elevator, on Bedford Avenue Brooklyn, 270 sq ft living room, equipped open kitchen, 2 bedrooms of 130 and 110 sq ft, original hardwood, double-pane windows, basement storage. Target buyer: young couple, first-time purchase. Highlight: the hardwood, the open kitchen, the dynamic neighborhood. Tone: professional and warm. Structure: strong hook in the first sentence, then strengths description, then neighborhood strengths. 200 words max. No spelling errors, no invented info.

78% of agents using AI use it for listing writing (RPR, 2025). No coincidence: it’s the use case with the best time-saved-to-quality ratio. To go further on the visuals that accompany these listings, see our method to improve your photos with AI.

Case 3: summarize a technical document in 2 minutes#

The problem: HOA meeting minutes, a 30-page energy disclosure, a purchase agreement with complicated clauses. These documents take hours to comb through, and often you don’t know exactly what to look for.

The AI solution: copy the text (or upload the PDF with ChatGPT Plus / Claude) and ask for a structured summary.

Here are HOA meeting minutes [paste text or upload PDF]. Summarize in 5 points: 1) assessments voted and their amounts, 2) impact on dues, 3) ongoing disputes or tension points, 4) important info for a potential buyer, 5) attention points for a real estate agent presenting this property.

What you get: a clear summary in 30 seconds. You can then ask precise questions: “Are there unusual clauses?”, “What’s the impact on a homeowner’s monthly budget?”. For this use case in particular, Claude (Anthropic) handles long documents better than ChatGPT. And the quantified property valuation in the next step, that’s our method to value a property with AI.

“That’s it? It’s that simple?”, the near-unanimous reaction to the first try. Yes, it’s that simple. The barrier was never technical.

Test now, before you close this#

Don’t read on before doing at least one of the three exercises above. Open ChatGPT. Copy a prompt. Adapt it to a real case from your week. Send. Time yourself. Compare with what you would have written alone.

The 3 beginner mistakes#

1. The vague prompt#

“Write a listing for my apartment” produces a generic unusable result. “Write a listing for a 730 sq ft 2-bed on Newbury Street, target young first-time buyer, warm tone, 200 words” produces a directly usable result. The level of detail in the prompt determines the quality of the result.

2. Using the output without proofreading#

AI can introduce inaccurate info if you haven’t provided enough. Always proofread, verify numbers, adjust tone if needed. The AI result is an excellent first draft, not a final document.

3. Giving up after a bad result#

If the first result isn’t good, reformulate. Tell the AI what’s wrong: “The tone is too formal, redo in warmer”, “It’s too long, cut to 150 words”. AI improves in the conversation.

The next step: personalize your AI#

Photo: Ketut Subiyanto · Pexels

The 3 use cases in this guide work with generic AI. But it doesn’t know your brokerage, your market, your tone, your usual templates. The next step is to build an AI assistant for your brokerage: with Claude Projects (~$20/month), you load your reference documents once (brokerage book, templates, market data), and Claude reuses them automatically in every conversation.

If you’re on a team and your broker is considering collective training, we’ve published a structured program to train your team on AI in 4 weeks, with a concrete deliverable each week.

Once these foundations are in place, you recover on average 3 to 6 hours per week. You reinvest them in the client relationship, where you really create value. The rest, AI handles.

Mistakes to avoid in your first 30 days#

The moment a real estate agent discovers AI is also the moment they pick up the worst habits. Once those habits set in, they slow progress for months and create the feeling that “it’s not as good as advertised.” Spotting the classic mistakes before falling into them saves the weeks that most agents waste fumbling. The learning curve for AI is fast when you avoid the traps — surprisingly slow when you string them together.

Four mistakes come up in 90% of failed starts I see:

• Testing in “one session” mode: opening ChatGPT, typing three prompts, closing the tab. The learning curve takes 15-20 minutes a day for two weeks, not a Saturday night marathon session.
• Copy-pasting without review: ChatGPT sometimes invents a school district, a square footage, or an HOA rule. Any output going to a client should be read line by line for the first weeks, until you spot errors at a glance.
• Keeping your prompts to yourself: a prompt that works for you will work for your teammates. Not sharing means depriving the whole brokerage of leverage that cost $0.
• Comparing AI to an expert human: ChatGPT will do worse than you on 10% of tasks, but 5x faster on the remaining 90%. The right benchmark is time/quality ratio, not absolute quality.

Agents who spend their first month avoiding these four pitfalls reach in 4 weeks what others take 6 months to reach. The difference isn’t IQ — it’s the starting protocol.

Why train the entire team#

The classic mistake is leaving training to individual initiative. Result: 2-3 motivated agents use AI, the rest continue as before. The competitive advantage stays marginal.

When a whole brokerage adopts AI consistently, the effect is different: listings have homogeneous quality, client emails are pro at every level, showing reports arrive the same evening. The external perception of the brokerage changes.

97% of US agents now use AI (WAV Group / Delta Media, January 2026). But only 23% have rebuilt at least one core workflow around AI (NAR 2026). The depth gap is the opportunity, but only for those who move now. For fundamentals and available building blocks, see generative AI applied to real estate.

The program in 4 weeks#

Each week has a precise goal, a main tool, and a concrete deliverable each agent must produce before moving on. The principle: one win per week, an immediately visible result that convinces even the most resistant.

Weekly duration: 1h30 of group training + 30 min of individual practice, 2 hours per week, no more.

Week 1: discovery#

The goal isn’t efficiency, it’s to lift the fear. Most agents who don’t use AI aren’t lazy, they just don’t know where to start.

3 simple tasks during the week#

• The tough email: give ChatGPT context (“follow up with an overpriced seller who hasn’t responded in 3 weeks”) and ask it to write the email. Compare with what you’d have written yourself
• The quick listing: give the info of a real property (square footage, floor, strengths, target) and ask for a 180-word listing
• The showing report: record a 2-minute voice memo, transcribe it, ask for a structured report, time it

Deliverable: each agent sends the broker a real email written with ChatGPT this week. No theory, a concrete result. For agents who want to get ahead, we’ve published a guide to get started with AI in real estate in 20 minutes.

Week 2: specialization#

Photo: fauxels · Pexels

This week introduces the assistant-that-knows-the-brokerage concept. That’s when AI stops being generalist and becomes specific to your firm.

Configure a Claude Project for the brokerage#

The broker creates a Claude Project (Pro, ~$20/month) following our guide to configure your Claude Projects. They load:

• The brokerage presentation book
• 3-5 email templates (follow-up, showing confirmation, report, objection responses)
• Local market data (price per sq ft, sale times, trends)
• Le style de communication (formel ou informel, tutoiement, vouvoiement)

Once configured, they show the team: “Now ask Claude to write a listing for this property, it already knows your context.” The quality difference between generic AI and AI that knows your brokerage is immediately visible. That’s the moment that convinces the last skeptics.

Deliverable: each agent produces a personalized template in the Claude Project. Their post-showing email template, in their tone, with their usual phrasing. The team has a consistent shared base.

Week 3: workflow integration#

This week connects AI to concrete moments in agents’ daily work: in the field, from the phone, between two showings.

3 workflows to integrate#

• Voice report: leaving a showing, the agent dictates a 90-second memo in mobile ChatGPT (mic icon), asks for a structured report. The email to the seller goes out within 10 minutes. Total time: 3 minutes
• Quick photo retouch: each agent tests Gemini Nano Banana or Adobe Firefly on an empty-room photo. Target under 2 minutes per photo. For the complete walkthrough, see our guide to improve your photos with AI
• Instagram content: each agent produces a weekly post with ChatGPT for the text. Paste the info of a property or tip, ask for hook + caption

Deliverable: each agent has integrated at least 2 of the 3 workflows into their real week, on a real listing or client, not as an exercise.

Week 4: autonomy#

Week 4 isn’t taught, it’s built individually. Each agent identifies the 3 tasks in their week where AI saves the most time, and systematizes them. For some, it’s listing writing. For others, seller appointment prep (including the data to value a property with AI). For others, document analysis.

The shared prompt library#

At the end of week 4, organize a 30-minute session where each agent shares their best prompt. Compile into a shared document: it’s the brokerage’s internal library, the most durable asset of the whole training.

Deliverable: each agent presents their personal AI workflow, the 3 tools used, the 3 prompts that save the most time, and the task they’ll never do again without AI.

The 5 classic mistakes to avoid#

Photo: Mikhail Nilov · Pexels

1. Train only volunteers: skeptics are the ones who lose most by not adopting. Training is mandatory for the whole team
2. Aiming for perfection in week 1: first results are imperfect, that’s normal. The goal is to trigger usage, not produce perfect content
3. Neglecting practice time: a demo without practice creates no habit. Each agent must use AI on a real case within 24 hours of each session
4. Too many tools at once: Week 1 ChatGPT, Week 2 Claude, Week 3 a visual tool. Not all in parallel
5. Not measuring: without measurement, no lasting conviction. Time tasks before and after AI, count post-showing emails sent in under 2 hours, note listing quality. Numbers speak louder than words

The 3 metrics that matter#

• Seller report time: how many hours between showing and email? Target: under 2 hours by program end
• First-pass listing acceptance rate: what proportion of AI listings is used without major changes? Target: over 80%
• Weekly time saved per agent: 3 hours per week per agent over 5 agents = 780 hours annually, equivalent to 4 months of work

A training that doesn’t change the numbers in 30 days didn’t happen. Measure, adjust, repeat.

Cost of training vs cost of not training#

Program cost: ~$20/month per agent (ChatGPT Plus or Claude Pro), 2 hours per week for 4 weeks, plus the broker’s investment to prepare templates and the initial Claude Project. Total: under $250 per agent and one month.

Cost of not training: continuing to lose listings to brokerages that respond faster, present better, and dedicate their time to client relationships while their AI handles the rest. Per Morgan Stanley, 37% of tasks are automatable. For a 5-agent brokerage at 40 h/week, that’s ~74 hours per week recovered, nearly 2 FTEs gained.

The ratio is so asymmetric the real question is no longer “do we train?”, it’s “who on the team leads the training?”.

The training format that actually sticks#

Most brokerages that try to train their team on AI default to a one-day workshop, sometimes external, sometimes internal. The result is predictable: the team is excited for two weeks, then reverts to old habits, and the brokerage owner concludes that “AI doesn’t work for our culture.” The format itself is the problem, not the team. AI training is fundamentally different from a software training because it’s not about learning a UI — it’s about building a new reflex of “what could I delegate to AI right now?” That reflex doesn’t get installed in eight hours of slides.

The format that produces measurable results is the opposite of a workshop. It’s a four-week sprint with three short sessions per week, each lasting 30 minutes maximum, focused on one concrete task that the agent must complete and report back on. Week one targets the easiest wins: rewriting a listing description with Claude, generating three variations of a Reel script, summarizing a long email thread. Each agent picks their own real-world task, runs it through AI, and shares the result with the team in a 5-minute show-and-tell. The brokerage owner reviews the outputs, gives feedback, and the cycle repeats.

By week four, the average agent has built a personal toolkit of 15-20 prompts they trust, has internalized the reflex of asking “is there a prompt for this?” before starting any task, and has produced enough wins to overcome the initial skepticism that always accompanies a tool change. Brokerages that follow this format report 70% adoption rates at week 4, versus the 20% adoption typical of workshop-format trainings. The difference isn’t the content. It’s the cadence: small reps, real tasks, shared learnings, sustained over a month rather than crammed into a day.

The brutal observation: a 2026 listing that looks like 2015#

Go to Zillow now. Search a 2-bed in Boston. Scroll 20 listings.

• How many have a 360° immersive tour? Zero
• How many have a video? Maybe two
• How many have well-framed professional photos? About 15%
• How many have a written description that sells, rather than a template? Almost none

Click on the brokerage profiles: 2012 sites, pale-blue headers, no blog, no Instagram (or a dead account, last post 2022). Typical outcome: a buyer lands on the competitor with a 360° tour + active Instagram. They call them. Not you. For the overall view of the shift, we’ve written how AI transforms the agent’s day-to-day in 2026.

The 5 symptoms of a lagging brokerage#

If you recognize 3 out of 5 symptoms, you’re actively losing listings.

1. Your photos are bad#

2015: iPhone 6, bad lighting. 2026: exactly the same photos. Properties with professional photos sell 15% faster and 8% higher. Cost: $500-$800 per property with a pro photographer. ROI amortized on the first accelerated sale.

2. Your descriptions are generic#

“Spacious 2-bed, well located, quiet, must see” in 2015 = same words in 2026. A real description is: emotional hook, clear advantages, quantified proof, call-to-action. ChatGPT produces one in 2 minutes, zero excuse.

3. You have no video#

Listings with video content generate up to +403% more leads. Not 40%, 403%. And “making a video” in 2026 = a smartphone + 2 minutes on CapCut. For the cinematic version, see cinematic real estate video.

4. You have no social media presence#

2015: Facebook for older folks. 2026: Instagram and TikTok = where your buyers live. +49% contacts on listings posted on social vs portals only.

5. Your site is frozen#

No blog, no updates, no 360° tour. Google demotes you each quarter. Your competitor who posts an article a month and publishes a tour per listing rises, you sink.

What it really costs#

Let’s be concrete. A 6-agent Boston brokerage, 180 active listings, consistent contact-to-sale conversion.

• Mediocre photos instead of pro → -30% clicks on the listing
• Generic descriptions → -25% additional clicks
• Zero video → -40% social shares
• Unoptimized site → -20% Google traffic

Cumulative effect: 30-35% less traffic to your listings. Instead of 100 contacts/month, you have ~65. Instead of 24 closed listings, you have ~16. 8 listings lost per month. At $700K average sale price and 3% commission, that’s ~$144,000 in evaporated commissions per year, ~5% of the revenue of a $3M brokerage.

What it costs to stay in 2015 is rarely a budget. It’s $144K of invisible commissions, going to the competitor.

The 5 blockers actually stopping you#

1. Organizational inertia#

“We’ve always done it this way, it works fine.” That’s a survivor strategy, not a winner’s. Your best agents leave for the competitor with better image, more contacts, more outlets. On the adoption case, we’ve published detailed data in our deep-dive why get started now.

2. Lack of internal skills#

You don’t know how to make a video, use Instagram, set up a chatbot. Solution: outsource to a freelance (~$600/month for 2 hours per week), or spend 8 hours learning the basics. No third option.

3. Perceived cost#

“360° tour = $6,000.” Wrong: ~$250-$350 with Kappn. “Pro photographer = $2,000.” Wrong: $500-$800. “ChatGPT for listings = complicated.” Wrong: free or $20/month. The perceived cost is ten times the reality.

4. No clear KPIs#

You don’t measure. You say “it works” without numbers. Solution: track for 30 days where contacts come from (Zillow, Instagram, direct call, referral), cost per contact, conversion rate by source. Truth surfaces in 4 weeks.

5. False excuses#

“Our clients are older, not on Instagram.” Clients aren’t looking for a brokerage, they’re looking for a property. 97% of searches start online, 35% outside portals. “We don’t have time.” Brokerages that succeed don’t have more. They take it. “It’s too complicated.” 2026 tools are simpler than 2015 tools.

What leading brokerages do#

Photo: Tima Miroshnichenko · Pexels

The top 20% of brokerages in their market follow the same principles, everywhere in the US.

1. Photos non-negotiable: pro photographer twice a month, ~25 photos per property
2. Custom description per property: ChatGPT or freelance, 15 minutes max, focus on benefits, proof, action
3. Video on 80% of properties: 360° immersive tour or 30-second Reel
4. Active social presence: 2-3 posts per week, mix of tours + neighborhood lifestyle + testimonials + tips
5. Updated site: 1 blog article per month, property-specific descriptions, qualifying chatbot
6. Monthly measurement: contact source, conversion rate by channel, budget adjustment

Measured impact on these brokerages: +30-40% contacts for an identical marketing budget. Properties sold 15-20% faster. Asking price hit 92% of the time vs 78% average. For the complete view of this transformation, see our pillar guide on AI in the real estate transaction.

The modernization plan, 4 steps in 6 weeks#

Step 1: pro photos and virtual staging (weeks 1-4)#

Action: pro photographer for your 10 best listings. Cost: $700 × 10 = ~$7,000 one-time. Impact: +15% clicks on these listings. In parallel, test virtual staging (Gemini Nano Banana, Adobe Firefly) on 3 empty-room photos: $0-$60 per property, fast decision on added value.

Step 2: punchy AI descriptions (weeks 2-3)#

Action: create a ChatGPT template prompt for your listings. Cost: $0 or $20/month (Plus). Time: 15 min per listing. Impact: +25% clicks, genuine, emotional, clear descriptions.

Step 3: immersive video standard (weeks 3-6)#

Action: 360° tour for your 15 best listings. Cost: ~$3,500 one-time + $700-$1,100/month for new listings. Impact: +40% contacts, -25% wasted showings (property already filtered on buyer side).

Step 4: social media presence (from week 2)#

Action: pro Instagram account + 2 posts per week. Cost: $0-$600/month (depending on freelance or in-house). Time: 2-3 hours per week. Impact: +15% contacts in months 1-2, +35-50% in months 4-6 when the algorithm has noticed you.

The ROI calculation of modernization#

Photo: Mikhail Nilov · Pexels

One-time investment: ~$12,000 (photos 7k + 360° tours 3.5k + virtual staging 1.5k).

Monthly investment: ~$1,300 (freelance 600 + new 360° tours 700 + ChatGPT 20).

Estimated gain: +30% contacts (from 100 to 130/month), +8% closed listings (from 16 to 17/month). At ~$15,000 average commission (3% on $500K), that’s +$15,000/month, or +$180,000/year.

Year 1 balance: -$27,600 in costs, +$180,000 in gains, or ~+$152,400 net. Year 2 and beyond balance: -$15,600 in costs, +$180,000 in gains, or ~+$164,400 net per year. These numbers are calculated for a standard 6-agent, 180-listing brokerage. Adapt to your size.

The cost of inaction#

Action costs ~$12K one-time and ~$1.3K/month. Inaction costs 8 listings lost per month, ~$144K/year of commissions walking out. The ratio is decisive.

In 2026, communicating like a modern brokerage isn’t a competitive advantage: it’s a condition for survival. Brokerages that move now will see results in 2-3 months. Those that wait until 2027 will arrive late: algorithms saturated, buyers acclimated elsewhere, agents gone to competitors.

You have the plan. You have the cost. You have the gain. The question is no longer “Can we afford to modernize?”, it’s “Can we afford to wait?”.

The state of portals in 2026#

Let’s start with a fact reassuring for Zillow: 97% of real estate searches start online (NAR). It’s a stable number since 2018. The internet isn’t in decline—it’s the post-internet phase.

But here’s the real number that matters: where online? In 2020, 85% of searches happened on Zillow, Realtor.com, Redfin, or equivalents. In 2023, it was 71%. In 2026, we’re around 63-67% per studies.

That means 35% of real estate searches no longer go through the portals. They go to Instagram, TikTok, Google Maps, ChatGPT, YouTube, even TikTok Shop. Zillow didn’t lose 50% traffic at once, but progressively loses 1-2% per month. And the erosion is accelerating.

The behavior that changed everything: under-35s#

Photo: Vlada Karpovich · Pexels

The data is clear: 63% of under-35 buyers start their search on social media, and only 28% start with a real estate portal.

1. The lack of social context on portals#

On Zillow, you see a photo, text, a price. On Instagram, you also see the life of the neighborhood: restaurants, streets, vibe. A friend comments: “I live around there, it’s nice.” Info no portal can provide.

2. Smarter algorithms than filters#

The TikTok algorithm knows you like small apartments with open kitchens, that you’re looking on the West Side, that you’re hesitating between the Upper West and Hell’s Kitchen. It shows you the right properties. Zillow does too, but you have to type 50 criteria and filter.

3. Instant sharing#

Your spouse finds a 1-bed on Instagram Stories. They send it in 2 seconds on WhatsApp. On Zillow, you have to copy the link, wait for the page to load, share. Too slow for a buyer in flash mode.

4. Video has won#

A 360° virtual tour on TikTok, you scroll by moving your phone. More immersive than a photo. The generation buying its first home grew up on YouTube, Netflix, TikTok. Static images bore them.

The numbers, no detour#

• Real estate Instagram: ~500 million searches/month tied to real estate (#realestate, #apartmentgoals, #househunting)
• Real estate TikTok: +400% year-over-year growth in real estate content since 2023
• Conversational search: 35% of under-25s use ChatGPT or Perplexity to search for a property
• Dark social: 56% of property shares happen via WhatsApp / Messenger / Telegram, invisible to portals

63% of offers without an in-person showing#

Here’s the number that should worry Zillow: 63% of buyers make an offer without visiting the property physically. Especially under-35s.

• The 360° immersive tour replaces the in-person showing. You see every corner, zoom, mentally measure ceiling heights. Why travel before making an offer?
• Investor buyers represent ~25% of US purchases. They’ve always bought at distance. AI just helps them do it faster and better.
• Timelines shrink: you find Monday, offer Tuesday, accepted Thursday. No time to mature, the offer has to fly or someone else jumps in.

Consequence for portals: classic visuals are no longer enough. Zillow displays 15 standard photos, the buyer wants immersive video and rich content. A brokerage with a Kappn tour (360°) on its site, shared on Instagram, receives +40% more contacts. On these dynamics, see our pillar guide on AI in the real estate transaction.

A concrete example: portal budget cut in half#

Photo: Tima Miroshnichenko · Pexels

An independent 12-agent brokerage in Boston spent $18,000/month on portals (Zillow Premier, Realtor.com, Redfin Partner, plus sponsored ads). $216,000 per year. In January 2025, they test a new strategy:

1. 70% of portal budget retained: $11,000/month (still present, but without premium packs)
2. Pro Instagram account + half-time community manager: $3,000/month
3. Immersive tours on each property (Kappn, ~$250/property): ~$1,200/month
4. Short TikTok videos: 1-2 per week (included with the CM)
5. Automated chatbot follow-ups: free or marginal cost

Total cost: $15,000/month. Savings: $3,000/month, or $36,000/year. And after 4 months:

• Contacts via portals: -8% (expected and accepted)
• Contacts via Instagram: +65% (immersive tours shared in Stories = viral effect)
• Contacts via TikTok: +35% (an AI staging video hit 120K views, 400 DMs)
• Contacts totaux : +22 %
• Average sale time: 16 days faster
• Prix atteint : stable ou +1 %

+22% contacts = 3-4 additional listings per agent per year. Across 8 selling agents, ~28 additional listings, or ~$50K in gross commissions. Net gain: ~$86K/year, while reducing the marketing budget. Zillow hasn’t vanished. But it’s no longer the only horse.

Conversational search: the new portals#

A repeating scenario in 2026: a buyer opens ChatGPT or Perplexity and asks “I want a 2-bed in Austin, under $400K, 10 minutes walk from a school, with parking. What do you suggest?”. AI synthesizes available data (MLS, indexed portals, brokerage sites, LinkedIn and Instagram content) and proposes 3 specific properties with details, valuation, market trends.

The buyer clicks on one, arrives at the brokerage site, sees the 360° immersive tour, requests an appointment, negotiates, signs. A brokerage with a rich website, a 360° tour, and structured content will be found by ChatGPT, Perplexity, and Google. A brokerage betting only on Zillow will stay invisible in this search. This doesn’t kill portals, but it dilutes their power: you’re no longer required to go through them to be discovered.

Zillow isn’t dead. Zillow lost the monopoly. You’re no longer hostage: you’re the decider.

The acquisition strategy in 2026#

Think omnichannel. Three simple steps.

Step 1: portal presence (60% of budget)#

You can’t ignore Zillow, Realtor.com, Redfin. Still 65% of entry points. Keep listings up to date, decent photos, complete brokerage profile. But stop the premium “super listing” and “boost positioning” packages—the ROI erodes.

Step 2: direct presence (30% of budget)#

• Rich website: dedicated page per property with 360° tour + AI estimate + pro photos
• Instagram and TikTok: 2-3 posts per week — see how to create real estate Instagram content that performs
• Google My Business: polished local presence
• LinkedIn: essential if you prospect investors or developers

Step 3: discovery automation (10% of budget)#

• Chatbot on your site, 24/7 qualification
• Clean Google indexing to be found by conversational AIs
• Shareable content with no friction on WhatsApp, Messenger, Stories

Is Zillow really in danger?#

No. Zillow isn’t in danger: it’s in transition. From “required marketplace” to “platform among others.” The likely scenario for the next 5 years:

1. Zillow remains dominant on desktop (~63% of traffic), but weakens on mobile (~47% of traffic now comes from social media and conversational search)
2. Zillow adapts: more 360° video integration, AI recommendations, Instagram-style feed
3. Large firms maintain big portal budgets. Small brokerages reduce and diversify. The gap widens
4. Innovations arrive elsewhere before Zillow. A new tool is born on TikTok or in a startup, Zillow copies 6 months later

Zillow will remain a key player. But never again the key player. On the case of brokerages that haven’t yet shifted, see our deep-dive on the need to modernize your brokerage’s communication before competitors do it for you.

Where to start#

You have a limited marketing budget. You can’t do everything. Here’s the priority order field-tested.

Priority 1: immersive video on your 5 best listings#

A 360° tour generates +40% contacts vs photos alone. The best signal-to-noise ratio in real estate marketing in 2026. Budget: ~$1,200/month, ROI visible in 4 weeks.

Priority 2: pro Instagram, 1 post per week#

No need for a full-time community manager. Someone who posts every Monday morning a 30-second immersive tour, that’s 2 hours per week. Cost: ~$600/month in freelance. Impact: +15% contacts in 2 months.

Priority 3: dedicated property page for hero listings#

A site dedicated to your best property generates more contacts than 10 properties on Zillow. Cost: $6-12K to build, amortized on 1-2 premium listings.

Priority 4: qualifying chatbot#

A chatbot asks the right questions (“Budget? Square footage? Location?”) and proposes 3 properties in response. Cost: $0-$600/month. 35% of visitors become leads, vs ~8% without chatbot.

The end of Zillow, is now#

Zillow isn’t dead. Zillow lost the monopoly. Same thing, except you’re no longer hostage. Brokerages that diversify now gain 20-30% more contacts, for an identical or reduced budget. Brokerages that wait will be blocked: everyone will have adopted Instagram and TikTok, they’ll arrive late, the algorithms will be saturated.

2026 is the year you start building your independence from Zillow. Not by abandoning it—by complementing it.

A new definition of the job#

There will never be a robot capable of selling a home. Not in 2026, not in 2030, not even in 2050. Selling is 80% human relationship: listening, trust, negotiation, psychology. But it’s also 20% repetitive tasks: writing listings, finding comps, admin follow-up, nurture, scheduling showings.

Here’s the shift AI triggers. The augmented real estate agent isn’t a replaced agent. It’s an agent whose 20% of low-value work is delegated to tools. They recover 6 hours a week. They use them to do what no AI can: build a real commercial relationship, negotiate finely, understand the seller’s unsaid, sense when a buyer hesitates.

In 2026, the agent who ignores AI looks like a travel agent who refuses the internet. They manage files by hand, write listings in two hours, make follow-up calls one by one. And their contact rates fall below those of an augmented agent. For the overall context, see how AI transforms the daily work of the job.

What AI takes over: low-value tasks#

Per Morgan Stanley, 37% of real estate tasks are automatable. They’re never the most important: they’re the most time-consuming.

1. Listing writing: a good listing in 3 minutes instead of 30
2. Finding comps: AI estimate in 30 seconds instead of 2 hours
3. Lead nurture: automated emails and SMS, no follow-up forgotten
4. Document synthesis: offers, contracts, purchase agreements analyzed in 2 minutes
5. Lead qualification: chatbots asking the right questions 24/7
6. Visual creation: virtual staging, photo retouch, 360° galleries
7. Admin tasks: data entry, calendar management, reminders
8. Targeted prospecting: potential seller lists based on MLS data
9. Transaction tracking: automated statuses, milestones tracked hands-free
10. Short trainings: the right info at the right moment via Claude assistant

None is strategic. Stacked together, they steal 10 hours a week from an average agent. On what AI actually changes in the remaining 90% of tasks, we’ve written a guide on what AI changes concretely.

What AI will never do#

Five human skills create real commercial value, and no model comes close.

1. Read non-verbal emotion#

A homeowner says: “I need to sell fast”. AI reads “fast.” You see the sweat on their forehead, hear the tone tick up a notch, understand it’s not a want, it’s an emergency. You adapt.

2. Negotiate under uncertainty#

AI can give the market price. It can’t see when a buyer bluffs, or when a seller is ready to give up 5% if you press here instead of there.

3. Build trust#

A seller is afraid of being scammed. AI can send a correct email. You spend an hour on the phone reassuring them, showing them your comparable sales, calling them three times to answer their questions. That’s what wins the listing.

4. Invent creative solutions#

A property won’t sell because it’s next to a highway. An augmented agent contacts a transport manager, finds an investor looking for a commercial space, and sells the property 8% higher. AI would have said: “bad location.”

5. Handle conflicts#

A buyer threatens to cancel due to a delay. A human agent finds a win-win. AI would have already closed the file.

Augmented prospecting#

Photo: fauxels · Pexels

Before: you called 50 people a week at random. Response rate: 4-5%. One sale every two weeks.

Now: AI analyzes public MLS data, identifies 200 potential sellers (6-year owners, bought before 2016, in the right neighborhood). You target them with a hyper-personalized SMS: “John, you bought your 2-bed on West Belden Avenue in 2016 at $510K. We help owners like you sell 25% faster.” AI sends the sequence 24/7. You call the 40 who respond.

Result: 12% response rate instead of 4%. 5 listings a week instead of 1. It’s not magic, it’s precision. And precision comes from AI.

Augmented valuation#

Before: you arrived at the seller’s with your printed list of 3 properties sold in the area. They’d say: “That’s low, my buddy was offering more.” Debate closed.

Now: you have an AI estimate based on 40 transactions on the street, over 24 months. You see the trends: “On Bedford Avenue, 2-beds sold between $440K and $480K between January and March. You’re on Smith Street, 50 yards closer to the F train, better exposure. We target $490K.” The seller feels it, it’s precise. Listing agreement signed.

More importantly: this estimate makes you arrive calm. Zero debate on valuation. You can talk marketing, timeline, minimum acceptable price. That’s where the listing is won.

Augmented communication#

Before: 200 listings, 200 text descriptions, zero video, zero social content. You posted on Zillow and waited.

Now: each property has an AI listing, virtual staging (3 versions), an immersive video, social content auto-published. You repurpose the video into 3-4 Reels clips.

Result: properties sold 25% faster, contacts +40%. And for client follow-up, automated nurture kicks in: D+3, D+7, D+14 if no response. No hot lead escapes because you “forgot to call.”

Augmented advice#

Before: you arrived at the appointment with a printed sheet and your know-how.

Now: you arrive with an automated summary of the 3 best properties for the profile, AI estimate of probable prices, MLS history, local trend, and rental simulation for investor profiles. You arrive as a real authority, not a salesperson pitching air.

The 3 new skills to develop#

If AI removes 20% of low-value work, it creates 3 new skills every agent must master to stay competitive.

1. Prompt engineering, knowing how to talk to AI#

“Write a listing” gives flat text. “Write a listing for a family 2-bed in Austin, premium tone, emphasis on quiet + near school + balcony” gives a text that sells. It’s a skill, you learn it in 2 hours, it saves 5 hours a week.

2. Reading data#

AI valuation gives a price. Why? Which comps? Are they really comparable? An augmented agent reads an MLS chart and sees: “Since January, 2-beds near a school sell 2% higher. My property is near a school. I refine my strategy.” That’s what creates the gap.

3. Tool orchestration#

You can’t use 20 tools at the same time. But you have to know the ecosystem: which tool for which need, how to connect them, how to automate them. An augmented agent knows: “For this prospecting, I use Apollo for the list, Make for automation, Calendly for the calendar, and a chatbot to qualify.” They set it up in an hour and it runs by itself.

To structure these learnings at the brokerage level, we’ve published a program to train your team on AI in a few weeks.

AI doesn’t replace agents. It replaces the agents who refuse to use it.

Lagging agent vs augmented agent: the gap widens#

The augmented agent saves 10 hours a week. They reinvest them to make more calls (better targeted), build stronger relationships, negotiate more finely, win more listings because their reputation rises.

Over a year: 25 more listings. At 3% average commission, that’s ~$75K more. The lagging agent watches and wonders why they’re losing listings to a competitor with “marketing support.” The competitor is just someone who augmented themselves with the tech.

The adoption plan, 3 months flat#

Photo: Andrea Piacquadio · Pexels

You can’t adopt everything at once. Here’s the progressive plan.

Month 1: automating critical tasks#

• Integrate ChatGPT or Claude for listings and emails writing
• Install a chatbot on your site
• Set up follow-up automation via Make

Month 2: data and valuation#

• Sign up for an AI valuation tool (Zillow Zestimate, HouseCanary, Plunk)
• Explore MLS data to refine prospecting
• Build a first targeted list of potential sellers

Month 3: content and immersive video#

• Launch one immersive video per listing (Kappn or equivalent)
• Create 1 social content per week (30-second Reel, improved listing)
• Measure impact: contacts, sale time, price achieved

By the end of month 3, you save ~6 hours per week, your contacts rise ~25%, your sales happen faster. At 12 months, the gap with a classic agent is structural.

From anxiety to power#

Many agents are afraid of AI. They think it will replace them. It’s understandable, and it’s wrong. AI replaces the agents who refuse to use it. It augments those who embrace it.

The future of the job isn’t an agent replaced by a robot. It’s an agent who spends two-thirds less on low-value tasks, and two-thirds more on what really matters: relationship, negotiation, trust.

The state of play in two numbers#

97% of US agents already use AI in their daily workflows (WAV Group / Delta Media, January 2026). But surface-level use only goes so far. The deep-adopters—those who’ve rebuilt at least one core workflow around AI—are roughly 23% per NAR’s 2026 technology survey. That’s the gap to close.

The gap between pioneers and laggards has a simple explanation: PropTech adoption is no longer limited to large firms. The tools have become accessible, affordable, essential. Brokerages that ignore these innovations in 2026 aren’t just behind—they’re progressively disappearing. For overall context, see how AI transforms the agent’s day-to-day.

Innovation 1: generative AI for writing#

Per Morgan Stanley, 37% of real estate tasks are automatable with generative AI. ChatGPT, Claude, and Gemini let you generate in seconds a complete listing, a prospecting email, an offer summary, or an admin document—genuinely adapted to the property and local market.

What it changes: 30 minutes instead of 3 hours to produce a month’s text. Prospecting emails become personalized instantly. Example: an agent imports a property’s raw description into Claude, gets three listing versions (standard, premium, creative) optimized for Zillow. Gain: 2 hours a week and better portal visibility.

Innovation 2: virtual home staging#

Listings with a virtual tour get +49% more contacts. At that level, virtual staging is no longer optional. AI analyzes a photo of an empty or cluttered room, then transforms it into staged versions: furniture added, colors optimized, brightness improved, virtual paint.

What it changes: empty properties stop being a handicap. Example: Gemini Nano Banana or Adobe Firefly transforms a photo of an empty living room into three moods (minimalist, cozy, family) in under a minute. The seller sees the potential, the accepted offer rises 8% on average.

Innovation 3: automated immersive video#

Photo: Mati Mango · Pexels

Properties shown with video sell 25% faster. Immersive video has gone from luxury content to sales standard. Tools like Kappn capture a property from photos, automatically generate an immersive video ready to publish on portals, the brokerage site, and social media. No video team, no complex editing.

What it changes: every property gets a professional video format. Buyers tour 24/7 from their couch. Wasted showings (in-person tour after video) drop by half. For the cinematic version, see our guide on cinematic real estate video.

Innovation 4: conversational search#

97% of real estate searches start online, but increasingly less on Zillow. A buyer asks ChatGPT or Perplexity: “I want a 2-bed in Austin, near transit, under $400K.” AI synthesizes public data (MLS, portals, social) and proposes specific properties, including from small brokerages.

What it changes: you’re no longer dependent on classic portals to be discovered. Your online presence, content, visuals are indexed by these AI engines. Example: a buyer searches “home with pool Austin $750K” on Perplexity. AI shows them 4 properties, including one your brokerage posted on Instagram two days ago. They click, watch the video, contact you. Portal cost: $0.

On this shift, we’ve written a complete deep-dive: real estate search is changing, and Zillow is no longer the required gateway.

Innovation 5: real-time AI valuation#

Property valuation is no longer a long, opaque process. Platforms like Zillow Zestimate, Redfin Estimate, HouseCanary, or Plunk analyze comparable sales, trends, and generate a reliable real-time estimate, with hyper-local data (neighborhood, street, sometimes building).

What it changes: you walk into a seller appointment with an AI-validated estimate. Zero debate on valuation. You save time and gain credibility. Example: a seller calls you at 7 PM panicked about a price. You scan the address, you have a reliable range in 30 seconds based on 12 transactions on the same street in 18 months. You sign the listing agreement on the phone.

Innovation 6: workflow automation#

Brokerages that manually manage email follow-ups, SMS, and calendar are losing listings due to wasted time. Make (formerly Integromat) or Zapier connect your tools and automate complete workflows.

Example: a buyer fills out a form on your site → automatic Calendly invite → if showing confirmed, SMS reminder 24h before → CRM add → email follow-up if no response after 3 days. Zero human intervention until the showing. You reclaim 100% of your time for the relationship.

Innovation 7: qualifying chatbots#

63% of buyers make an offer without a physical showing. Your online visitors expect an immediate response, not a call the next day. An AI chatbot asks the key questions in real time (budget, square footage, location, timing) and qualifies the lead in 3 minutes.

Example: a visitor arrives on your site at 10 PM, clicks “Find my home.” The chatbot chats for 2 minutes, sends 3 proposals, and only unlocks contact info if the user “bites” on a property. In the morning, you have 5 hot leads, not 20 noise.

Innovation 8: MLS data for prospecting#

Photo: David McBee · Pexels

MLS and public records data are no longer a secret kept by a few expensive industry tools. ATTOM Data, Reonomy, Estated, or even Python scripts analyze public transactions by area, identify potential listing leads (recent sellers, 10+ year owners), and provide ultra-targeted prospecting lists.

Example: you target Austin condo owners who bought between 2015 and 2018, statistically likely to sell in 2026. You find 1,200 people. Personalized SMS sent: “David, you bought on East 6th in 2016 at $260K. We help owners like you sell 12% faster.” Average response rate: 8%, ~96 potential listings on the table.

Innovation 9: AI prospecting#

Prospecting is ~70% of a brokerage’s work, and ~70% of prospecting is repetitive and automatable. Apollo.io, ZoomInfo, or HouseCanary automate prospect search (owners, investors, builders), qualification, and first email contact.

Example: filter “multifamily owners in Brooklyn, 5+ units, bought before 2010.” The tool identifies 340 targets, sends an automated email sequence. You manually handle the 40 hot replies, that’s it.

Innovation 10: dark social and TikTok#

Real estate searches no longer start on portals. They start on Instagram, TikTok, and even TikTok Shop. Dark social (WhatsApp, Messenger, Telegram, Instagram DM) amplifies everything: a friend sends a tour video to three friends, without your knowing.

Example: you publish a 30-second Reel “Before/after virtual staging.” A local TikTok creator reshares it. 400,000 views. 200 DMs from young regional buyers. Cost: $0. ROI: infinite, provided you have the system to handle leads.

70% of brokerages communicate like it’s 2015. The other 30% take their listings and their buyers.

The next step for your brokerage#

The 2026 US PropTech landscape offers a simple opportunity: 70% of brokerages have integrated none of these 10 innovations. This majority communicates, prospects, and sells exactly like it’s 2015.

Those that have adopted three or four of these tools already see results: +40% inquiries, -25% sale time, -30% admin work. The question isn’t whether you can afford to adopt PropTech—it’s whether you can afford to ignore it.

For the overview and recommended adoption sequence (what to integrate first, second), see our pillar guide on AI in the real estate transaction.

Start small: one immersive video (Kappn) on a listing, a chatbot on your site, a follow-up automation in Make. Measure at 3 months. You’ll see the difference.

How to pick your first 3 PropTech solutions#

With the US PropTech ecosystem now exceeding 2,000 active startups, the question isn’t “which innovation to test” but “how to avoid spreading thin.” Brokerages that succeed in the PropTech transition don’t run pilots on ten tools in parallel — they focus on three precise fronts that together cover over 80% of the possible productivity gains. Choosing those three fronts depends on your current situation, but the decision framework stays the same.

Three questions frame your choice without taking on risk:

• Where do you lose the most time this week? If the answer is “data entry and reporting,” start with an automation tool like Make or Zapier. If it’s “listing copy and content creation,” start with Claude Projects. If it’s “lead qualification,” start with a conversational AI agent.
• What do your clients complain about most? Slow response time, lack of updates, incomplete files — every client pain point maps to a specific PropTech category. Invest there first.
• What data do you already have? A brokerage with 500 historic listings well-structured can build a predictive AI valuation engine. A brokerage with a messy CRM must first clean its data before layering AI on top.

Brokerages that answer these three questions before choosing gain 6 to 12 months on those who jump on the first slick demo. PropTech is a marathon of combinations, not a sprint on the trendiest tool.

The problem: an empty room loses 30% of its appeal#

A brutal fact: a property with visual quality sells 25% faster on average. For a buyer, “quality” doesn’t mean a $5,000 physical staging with decorators and rented furniture. It means a photo that shows the room’s potential.

That’s exactly the snag. You have an empty bedroom, a furniture-less living room, a basic kitchen. In the photo, it looks “lifeless.” The buyer doesn’t project themselves. Consequences:

• Fewer listing clicks
• Fewer showings requested
• More price negotiation (“empty room = renovation expected”)

For overall visual quality before even staging, see our method to improve your photos with AI: brightness, color balance, uniform sky. That’s the prerequisite.

Now imagine the same room with a bed, a dresser, some decor. Suddenly, the buyer sees where their sofa goes, they project living there. That’s the goal of AI virtual staging.

Virtual vs physical staging: the cost disruption#

Classic physical staging is:

• Rent furniture: $400-$1,200 for a home
• Bring in decorators: 1-2 days of work
• Photograph the staged property
• Remove furniture after the sale

Total cost: $5,000 to $15,000 per property. Viable for an $800K home, completely out of reach for a $250K studio.

AI virtual staging is:

• Upload an empty-room photo
• AI generates multiple furnished versions (bed, sofa, minimalist or warm decor)
• Download the enhanced image
• Publish with “virtually staged” disclosure

Cost: free to $60 per property. Time: 2-3 minutes per room.

Measurable results: +25% sale speed#

Photo: Vecislavas Popa · Pexels

Properties shown with quality visuals (physical staging, video, AI) sell 25% faster. On a $750K home:

• Without staging: 90 days on average
• With AI staging: 67 days

23 days saved—a commission cashed three weeks earlier, and a listing freed for the next. On inquiries, listings with enriched visuals (staging + video) receive +49% more contacts. More contacts + faster sale = more properties sold with the same prospecting effort.

Gemini Nano Banana tutorial, step by step#

Gemini Nano Banana is Google’s AI image model, free in the Gemini app, that lets you virtually furnish any empty room in seconds via a simple prompt.

Step 1: upload your photo#

1. Download the Gemini app (iOS / Android) or go to gemini.google.com
2. Click “Create a staging”
3. Upload your empty-room photo (JPG or PNG, 1 MB minimum)

Step 2: choose a furnishing style#

• Minimalist: clean furniture, white/gray, zen mood
• Warm: wood, warm colors, cozy
• Modern: clean design, gray/black, very current
• Boho: colors, plants, boho-chic
• Classic: timeless style

Tip: generate 3 versions of the same room, in different styles. Not all buyers share the same taste, and these variants also serve you on social media.

Step 3: adjust the settings#

• Furnishing budget: economy, standard, premium (Gemini adjusts furniture quality through your prompt)
• Light direction: natural lighting improvement
• Background blur: useful if the back wall is ugly

Step 4: generate and download#

Click “Generate.” AI takes 30-60 seconds. You get the retouched photo back in high resolution.

Step 5: publish with required disclosure#

On Zillow or Realtor.com, place the Gemini Nano Banana photo next to the real photo, and add a caption: “Virtually staged. Computer-generated image.”. Transparent, legal, and the buyer sees the potential without being misled.

Adobe Firefly tutorial, the photorealistic option#

Adobe Firefly is more powerful than Gemini Nano Banana on photorealistic precision. Ideal for premium brokerages that already have the Adobe suite.

Advantages#

• Fine furniture control: “light wood bed, gray pillow, no headboard”
• Photo generation, not simple retouch: AI really reinvents the room
• Multiple styles in parallel: 5 versions generated at once
• High-res export: print possible, easy later retouching

Pricing#

• Free: 1 generation per month (for testing)
• Pro: ~$18/month for 15 generations per month (per room, not per home)
• Brokerage: ~$60/month for unlimited generations

For a small brokerage (under 20 properties a month), the Pro plan is more than enough.

DALL-E 3 tutorial, via ChatGPT#

If you already have ChatGPT Plus, DALL-E 3 is the logical choice: integrated into ChatGPT, natural conversation to adjust renders. Limits: you have to clearly describe the expected render, standard resolution, 30-60 second generation time. Perfect if you already use ChatGPT daily.

Quick comparison of the three tools#

Recommendation:

• You’re starting: Gemini Nano Banana, free in the Gemini app
• 10+ properties per month: Adobe Firefly, best quality/price ratio
• High-end target: Midjourney, premium results for premium clients

Required disclosure, non-negotiable#

In the US, NAR’s Code of Ethics and most state regulations require you to disclose that the photo is virtually staged. Not optional. Three possible placements:

• Below the photo: “Virtually staged”
• At the bottom of the listing: “Photos include virtually staged images”
• Ideally, both

If you don’t disclose: risk of buyer complaint, loss of trust, reputation drop. AI does beautiful work, but transparency is non-negotiable.

AI furnishes the room. The “virtually staged” disclosure keeps trust. Both go hand in hand.

Complete workflow, from empty photo to listing#

Photo: Max Vakhtbovycn · Pexels

Day 1: tour and photos#

Photograph each empty room. Staging priorities:

1. Bedrooms: show where the bed goes
2. Living room: that’s where the buyer projects themselves
3. Kitchen: less priority, often visible “as is”
4. Office or secondary room: optional

Day 2: AI staging (30 minutes)#

1. Upload the key photos to Gemini Nano Banana or Adobe Firefly
2. Generate 2-3 versions per room, different styles
3. Download the best images
4. Integrate AI photos and real photos in your listing

Day 3: publication#

Publish on Zillow / Realtor.com with a mix of real photo + AI photo, “virtually staged” disclosure on each. Average result: +49% more inquiries on this listing.

Total time: 30 minutes of AI staging for a whole home. Cost: $0 to $60. Gain: 25% sale speed. The ROI is even sharper because the commission lands several weeks earlier.

To go further, from photo to immersive video#

You’ve got AI-staged photos. The next step: turning these images into a real estate video. That’s exactly Kappn’s promise: start from your photos (AI-staged or real) and generate a complete immersive video, with motion, narration, music. Video ready for Reels, TikTok, portals—no editing required.

For the complete picture of AI in the real estate transaction (from photo to closing), see our pillar guide on AI in the real estate transaction. And to push the render toward cinematic (drone, color grade, scripted narration), see cinematic real estate video.

The winning combo: Gemini Nano Banana or Adobe Firefly for staged photos + Kappn for video. Complete home presentation in 1 hour of work, $5-$120 total cost. And a property that sells 25% faster, on average.

Common mistakes that ruin a staged photo#

Virtual staging is a multiplier when it works and a liability when it doesn’t. The difference between a photo that pulls buyers in and one that flags as “fake” comes down to a handful of execution mistakes that most agents never even realize they’re making. After reviewing thousands of staged listings on Zillow and Realtor.com, three categories of error account for 90% of bad outputs. Avoiding them is not a matter of better tools — it’s a matter of better instructions.

The first category is furniture proportion errors. The AI doesn’t always know the actual size of the room, so a generated sofa can end up too large, blocking circulation paths, or too small, making the room look cavernous. The fix is to include scale references in your prompt: mention the approximate square footage, point out the door and window positions, and request furniture sized for “a typical 14×16 foot living room.” The model adjusts its output accordingly, and the result reads as natural rather than cartoonish.

The second category is style mismatch with the property. Generating a hyper-modern Scandinavian living room in a 1920s Victorian creates dissonance that buyers detect instantly, even if they can’t articulate why. The right approach is to match the staging style to the architectural period and the target buyer demographic. A craftsman bungalow in Portland and a high-rise condo in Miami need radically different staging vocabularies, and the prompt should specify both the architectural context and the buyer profile to align.

The third category is over-staging. Adding eight pieces of furniture, two rugs, three plants, four pieces of art and an accent chair turns the photo into a furniture catalog rather than a home. Buyers want to project their own life into a space, not feel like they’re walking into a showroom. The rule of thumb that works in 95% of cases: stage with three pieces of furniture maximum, one rug, one plant, one piece of art per room. The result reads as a real home — the only outcome that converts to showings.

Why Instagram has become essential#

The numbers don’t lie:

• 97% of real estate searches start online
• Listings with video content generate up to +403% more leads
• 94% of listings still don’t have a virtual tour or video

Two implications. First, you’re behind if you only position on portals (Zillow, Realtor.com). Second, Instagram is the channel where you capture these buyers before your competitors, because most of them aren’t really there yet.

And it’s not just visibility. Real estate Reels generate direct inquiries, without going through a portal. Those inquiries are your real leads, sometimes better qualified than portal leads. For more on multi-channel acquisition, see our guide on how to prospect with AI.

The 4 formats that work in real estate#

Before we talk tools, you need to know what to post. Four formats dominate in 2026.

1. Quick tours (30-60 seconds)#

Quickly film a property, or a key room, with your phone. Dynamic music, AI-generated captions, and you publish. Simple, effective.

Typical hook: “A 200 sq ft kitchen with island? Rare in this neighborhood. More in DM.”

2. Tips and advice#

“5 mistakes to avoid when touring a property”, “How to negotiate the price”, “Why HOA minutes change everything”. Buyers aren’t just looking for properties, they’re looking for advice. This is the format that positions you as expert, not seller.

3. Market commentary#

“Mortgage rates are dropping: what changes for you”, “2026 market: neighborhoods to watch”. These take little production (text, a few numbers, a clean visual), but generate tons of shares.

4. Before / after (staging, renovation)#

Before: empty or dated room. After: AI-staged or renovated room. Highly visual, stops the scroll, and proves the property has potential. This carousel type performs particularly well on Instagram.

AI for your text and hooks: ChatGPT and Claude#

Photo: RDNE Stock project · Pexels

A good hook is half of Instagram success. And writing 8 hooks a month takes time. AI is instant.

With ChatGPT, for quick hooks#

I’m a real estate agent. Give me 5 hooks for real estate Reels (30-40 characters max) on these topics: 1) tour of a 2-bed in South End Boston, 2) negotiation tip, 3) 2026 market, 4) before/after staging, 5) buyer mistakes. Direct tone, no jargon.

Result in 10 seconds. Adapt one or two versions, and you’re set for the week.

With Claude, for longer content#

I’m a real estate agent in [your area]. Write an Instagram caption for a Reel on the 2026 real estate market in our area. 120 characters max. Professional but cool tone. Hook in the first line.

Claude excels at developed captions, market explanations, breakdowns where tone matters as much as info.

Mistakes to avoid in your text#

• Generic hooks: “Check out this gorgeous property” triggers nothing. Prefer: “At this price, this home won’t last 15 days. Why?”
• Text too long: Instagram rewards punchy, not blabber
• No CTA: always end with “More in DM” or an open question

AI for captions and editing: CapCut, Opus Clip, Submagic#

You’ve filmed a tour. Now you have to make it engaging. AI tools do that in 2-3 minutes.

CapCut (free)#

• Upload your video, CapCut auto-generates captions automatically
• Add music (free integrated library)
• Export in Reel format
• Temps total : 3 minutes

Opus Clip (free, paid options)#

• Perfect for cutting a long video into short Reels
• Upload 5 minutes of a tour, Opus generates 3-4 Reels of 30-60 seconds
• Each clip has its captions and music
• Ideal if you have long videos to recycle

Submagic (~$10/month)#

• Very fast captions, automatic repositioning, premium effects
• If you publish more than 10 Reels a month, the subscription is worth it
• Direct export in Reel format

Editorial calendar template: 9 posts per month#

Photo: Karolina Grabowska · Pexels

Here’s a concrete plan to start without feeling overwhelmed. 2-3 posts per week is the cadence that maintains visibility without saturating your audience.

Volume tip: prepare all your hooks and captions in one monthly session (1 hour with ChatGPT is enough for 4 weeks). Then post on the fly, without having to think.

5 ready-to-copy prompts to generate your content#

Prompt 1: monthly plan#

I’m a real estate agent in [YOUR AREA]. I want to create 8 Instagram posts per month. Give me an editorial calendar: 4 quick tours, 2 tip posts, 2 market breakdowns. For each post, propose a hook (30 characters max), the visual type, and the recommended AI tool. Table format.

Prompt 2: hook for tour Reel#

I have to make a Reel for this home: 1,600 sq ft, 4 beds, 3,200 sq ft yard, $850K in [NEIGHBORHOOD]. The market is tight, properties like this sell fast. Give me 5 different hooks (30-40 characters max, no emoji). Direct, engaging tone, not salesy.

Prompt 3: long caption for carousel#

I’m making a before/after carousel of a renovated home (kitchen, living room, bedroom). Write an Instagram caption 150-200 characters max. Professional but accessible tone. Start with a strong hook. End with a CTA (DM, appointment).

Prompt 4: market commentary#

I’m a real estate agent in [YOUR AREA]. Mortgage rates are dropping and that changes everything for buyers in 2026. Write a script for a Reel (60 seconds of narration): 1) the rate context, 2) what it changes for a typical buyer, 3) why now is the time to act. Factual, punchy tone.

Prompt 5: tips series#

I’m a real estate agent. Give me a list of 10 short tips (2-3 sentences max each) on: mistakes to avoid in tours, negotiation, neighborhood choice. Each tip should be a standalone Instagram post (Reel caption).

One hour of prompts per month. Four weeks of content banked. The ratio you were waiting for.

The mistakes that kill engagement#

1. Posting only static photos#

A classic real estate photo (facade, empty living room) is dead on Instagram. Reels generate 3-5x more engagement. Film, cut into clips, add captions.

2. Only talking about your listings#

If 100% of your posts are “Here’s a property for sale,” you’re selling, not communicating. Aim for 50% properties and 50% advice + commentary.

3. Ignoring DMs#

Someone comments “I like this neighborhood, any advice?” and you take 4 days to respond—it’s dead. Respond in under 2 hours. A converted DM is often an appointment.

4. Posting at random times#

3 AM: zero visibility. Holidays: no one scrolls. Aim for Monday to Friday, between 12 PM and 6 PM for most of your posts.

5. Reels that are too long#

A complete 3-minute tour: no one watches to the end. Instagram rewards videos under 60 seconds. If you have a long tour, cut it into 3-4 Reels.

To go further#

Once your rhythm is set (2-3 posts per week), the next step is connecting Instagram to your sales funnel: where does the DM go, who handles it, how does it become an appointment. See our guide on digitizing your sales process to avoid losing social-media-won leads. And before the first showing, prep your CMA with the method to value a property with AI based on MLS data.