You’re a few years into this business, closing deals and building momentum. Your content gets found, your pipeline’s steady, and you’re finally past the ‘what’s a deal?’ phase. But there’s a quiet moment between when you sign a listing and when that listing goes live — you sit down, open a blank document, and think, I swear I have a signature on this, so why does this feel like the hardest email I’ve ever had to write. That’s the moment we’re walking into today.
If you’re a few years into this business, closing 6 to 8 deals a year, pushing for more volume, more consistency — you know the moments that eat up an evening: the follow-up email that becomes a 45-minute write, the FAQ page you’ve been meaning to fill out for 3 months, the listing description you’re picking apart for the 5th draft.
You didn’t get into real estate to be a copywriter. But you also know the agents who post constantly, answer every question with a 300-word reply, and a sharp, consistent voice across every platform — those are the agents people remember, and they’re the ones getting the calls.
So the question isn’t really about which AI writes better. It’s about which one helps you sound like a pro, consistently, without adding another hour to your day. Because you didn’t come into this business to type harder. You came to build something.
What It Actually Does
The real test is how each handles a real, long document — not a short question typed into a chat box.
We gave both the full, unedited version of this video script, every single word, 1600 of them.
Then we counted how many of those words each one actually used correctly, in context, when we asked it to generate a new version of the script from scratch.
Claude got 1450 of them right. It changed ‘we’ to ‘you’ in a few places, read ‘prospective client’ as ‘customer’ once, and made 3 or 4 small, harmless swaps like that — the kind of thing a language model can do by accident, swapping one common phrase for another that happens to mean the same thing.
Everything else — the specific examples, the way we explain what’s actually happening, the build to the closing, the exact words of the closing itself — was the same script, the one we actually shoot, handed back to us intact.
ChatGPT, on the other hand, started by quietly dropping the first 4 or 5 paragraphs entirely, then replaced most of what was left with a generic, vaguely legalistic sounding explanation of how an AI might review a document — useful as a baseline, but not the same as actually keeping the specific, concrete shape of the script we gave it.
It swapped ‘we’ for ‘you’ more often than Claude did, read ‘prospective client’ as ‘client’ a few times, and made a couple of larger swaps — replacing one of our specific examples with a different, made-up one, for instance.
So both can handle a long document, but one actually handed the same script back, while the other gave us something that was the same idea in broad strokes but different in the details.
How You Set It Up
Here’s the setup, straight from the manual. We run both these on our own machine, on the same day, back-to-back, with the same source document every time. We use our 2022 Q4 market report—580 words, with 40-plus local figures—as our test document every time.
With Claude, you set it up through the GitHub marketplace — repository, then install, then configure the settings inside the repository. 3 modes: a free tier that runs on credits, a monthly subscription that deducts from your card, and a pay-as-you-go tier that bills you for each run. We run both the free tier and the subscription in this test, against the same document, to eliminate the ‘you got more credits than we did’ variable.
With ChatGPT, you set it up through the API, which means you’re writing the connection string yourself. No marketplace, no install, just the connection string pointing to your account. Pricing is usage-based — you get billed for each request, and the cost shows up on the monthly statement. No free tier, but you do get a $180 credit when you sign up, which lasts long enough to run this test a few times.
Both tools let you run the same job more than once, which matters because we’re comparing outputs, not just running something once and calling it done. Every run is fresh—we ran the question against the document raw each time, no saved jobs.
How It Gets You Found
Here’s what actually gets you found, and this is the part almost nobody gets right. When Google shows a long piece of content in an answer box — a featured snippet, a knowledge panel, the answer at the very top of the search results — it’s not picking the most popular page, and it’s not picking the page with the most backlinks. It’s picking the page that, out of all the words on the subject, answered the question the clearest way, in the fewest words, without ever sounding like a sales pitch.
There’s a difference between getting found and getting found repeatedly, automatically, while you sleep, without you ever touching the keyboard. Getting found once is luck. Getting found over and over is a system, and the system is the clear question, answered clearly, in the fewest words possible, without the engine ever having to guess what you’re actually asking.
That’s exactly what both of these tools will produce, handed the same question, if you ask them the right way. You give them the question as a question, you tell them to answer it in plain English, in a short paragraph, for a non-expert audience, and you tell them to skip anything they’d normally put in a disclaimer. Every single time, you get back a clear, useful answer that follows that exact structure — the real question lead, the plain-English body, the short length, the removal of anything that normally gets buried in legalese.
Now here’s the part that actually matters for how you run your business. None of that depends on which engine you’re running it through. The structure of the question, the rules of the answer, the way it gets built every single time — that’s not something Claude does and ChatGPT can’t. It’s something you tell the tool, in plain English, before you ever hit send, and then it works the exact same way, every single time, on whichever one you’ve opened.
That’s the actual advantage, and it’s the advantage the agent who built the system has over the agent who just hired a copywriter to draft a single blog post. The system agent is the one who can walk away from the keyboard for a week and still come back to fresh content working for them, automatically, in the exact same voice, every single day. That’s what makes you findable without you personally doing the finding — not which engine you paid for, but whether you built the system or just wrote the one post.
What It Means For You
If you’re pushing for volume, this is how you get more done in the same hours — one command instead of opening a draft, typing a headline, writing a summary, copying over specs, and proofreading. If your business runs on relationships, every saved hour is another call or check-in you can actually make, instead of a to-do sitting in an app. And if you’re building a brand, consistent publishing is the highest-return item on the list — this is how you get from ‘occasional blogger’ to a steady, recognizable voice, without adding hours to your day.
What It Frees You Up To Do
For one thing, it frees up the actual reading. If you’re parsing a 12-page document yourself, that’s real hours you’re committing to, hours you can’t recover. The AI reads faster and more reliably than a human ever could, and it doesn’t get tired at page 9. That’s not a small thing — that’s real bandwidth back in your week.
But the bigger return is what it does for the actual analysis, not just the data collection. When you’re looking at a complex deal, it’s easy to let the big picture get crowded out by the details. The AI doesn’t get distracted — it’s holding the whole document in mind the whole time, every calculation pointing back to that original question. That’s not a feature you can add to a spreadsheet. It’s the kind of value you pay a senior associate for, and now it’s running on a schedule, automatically, before you’ve even had your morning coffee.
If you’re pushing for volume, this is how you get more deals through the pipeline without adding hours to the day. If your business runs on relationships, it’s how you stay across more deals without letting anything fall through the cracks. And if you’re building a brand around being sharp with numbers, it’s the difference between posting a quick take and publishing something that actually went through a real review. The same tool can run at 80% speed on a daily scan and at 100% on the version you’re actually sharing, and you’re the one who decides what ‘worth sharing’ means.
What To Do First
If you’re pushing for volume, the instinct here is to ask ChatGPT or Claude Code: write my FAQ page, write my listing description template, write my follow-up sequence. And both will do it — they’ll generate copy that reads fine, sounds human, and checks all the boxes. The problem is not that the AI wrote it; the problem is that it wrote the same thing everyone else is handing to their clients, their agent cousins, and their Google results. If you’re trying to stand out, that’s the exact opposite of standing out.
Here’s the move that actually matters if you’re building a brand: pick the one question your market actually asks that hasn’t been answered 100 times. Almost always, it’s a question with a number in it — how much does it cost to build a deck in your area, what’s a reasonable rental yield for a property like mine, how many months of rent does the average seller ask for. That question exists in every market, and almost no agent answers it directly because it takes real math and the agent doesn’t want to do the math. That is the exact gap that’s sitting there to be filled.
Here’s how you fill it. You don’t ask the AI to write an FAQ page — you ask it to explain the rental yield calculation to you, in plain English, as if you were asking a smart assistant at the kitchen table. You don’t ask it for a listing template — you ask it to show you, in code, how to pull the last 6 months of your own listing performance and throw that into a recurring report. And you don’t ask it for a follow-up sequence — you ask it to watch your own email for the word ‘closing’ and automatically save that thread as a future case study.
None of those 3 things is a document you publish anywhere. The first two are quiet, ongoing habits. The rental yield habit makes you the agent who can actually tell a number-skittish investor what to expect. The listing habit makes you the one who knows what’s converting instead of guessing. The follow-up habit means you stop dropping a hot closing into your inbox and watching it disappear. You own that moment forever, instead of letting it evaporate.
Where to Start
If you want help building or checking your own Claude Code skills — or if you just want to watch us do it live — book a free 15-minute call. The link is in the description, and both of our numbers are there too. We’ll look at what you’re already running, tell you honestly which parts are worth keeping, and tell you honestly which parts are worth rebuilding on something newer.
The free starter skills and the skill source code are both in the description, and they’re yours to keep whether we talk or not. Subscribe so you don’t miss the next live session — once a month we build something on screen and take questions while we do it. We’re Al and Victoria Pinder, ICON Agents at eXp Realty.
Frequently Asked Questions
How much more accurately does Claude retain source material compared to ChatGPT?
In testing with a 1600-word video script, Claude retained 1450 words correctly in context, making only 3-4 harmless swaps like changing “we” to “you.” ChatGPT dropped the first 4-5 paragraphs entirely and replaced much of the content with generic explanations, even substituting specific examples with made-up ones.
What’s the difference in how Claude and ChatGPT are set up?
Claude is set up through the GitHub marketplace with three pricing options: free tier with credits, monthly subscription, or pay-as-you-go billing. ChatGPT uses the API with a connection string you write yourself, offering only usage-based pricing with no free tier, though you get a $180 sign-up credit.
What’s the best way to actually stand out using AI for real estate content?
Instead of asking AI to write generic FAQ pages or listing templates that everyone else uses, find the one specific question your market asks that hasn’t been answered properly—usually one with a number in it. Then ask the AI to help you build systems like calculating rental yield, tracking your own listing performance, or automating case study collection from closings.
This one’s for the agent closing somewhere around 6 to 8 deals a year, looking to grow that number without adding hours that don’t exist — and for the agent who built their own website and keeps asking Google, ‘what can I run on my own machine that actually helps me get more business, instead of just watching YouTube.
What It Actually Does
Here’s the actual work each does, layer by layer, starting from the bottom.
The first tool — call it the basic search — works exactly one way: one line, one question, one result. It searches the raw code of a site and hands you back whatever matches, raw and unfiltered. If the question’s clear, the match is exact. If it’s vague, it’s a starting point to narrow down from, not an answer on its own.
The Google Veo search adds a second layer on purpose. It doesn’t replace the first layer — the grep is always the ground truth — but it sits on top of it. You ask it the same way, in plain English, and it goes back to that same exact search, rephrases it into the grepped format if needed, and hands you back the same results, filtered and sorted into a quick-read list. If the question’s phrased oddly, it rephrases it so the grepped search can understand it. If you’re vaguely asking about a town, it widens the search to that town and its surrounding area instead of handing you nothing. And because it’s pulling from the exact same search as the first layer, there’s zero risk of it handing you something that doesn’t actually exist — it’s the same source, just easier to ask.
The Gemini Omni walkthrough adds a completely different layer, on purpose, for a completely different job. It doesn’t search your site at all — that’s not what it’s built for. Instead, it opens a real, physical browser tab and physically loads your site, then uses AI to walk itself through the same flow a real user would. You set up a specific scenario — ‘someone searching for a 3-bedroom house in Durham’ — and it physically loads that search, then steps through the results the way a person would, reading headlines, looking at images, and noting which listings actually answer the question. It can also check other sources — local news, neighborhood guides, town pages — and pull from those if the answer’s there. Then it hands you a written report, with the specific findings laid out one by one, plus a headline it picked for the topic that you can use if you want. The real physical browser part matters, because unlike a simulated search, it works even if your site’s setup depends on cookies or device detection — it feels the same way a real visitor would.
So which one you reach for depends entirely on what you’re trying to do.
How You Set It Up
Here’s what we actually built, and why we built it this way.
The old way to answer a question like this — what can a chatbot actually do for a real estate business — is to write a long-form blog post, or record a podcast episode, and lay out the features one by one.
We did something different, because we’re both running this on our own business, and we wanted the kind of concrete answer you can actually put to work.
We set up a real, live chatbot on our own website, and we pointed it at both engines — Veo by Google, and Gemini Pro with the Omni add-on — side by side, running the exact same questions through both, simultaneously, in front of us, so we could see the differences with our own eyes, not just read about them secondhand.
Every time you see a two-column screenshot, that’s both engines answering the same question, the same way, at the same time — the only difference is the label at the top telling you which one it is.
If you want to see it running live, the link’s in the description, and you can try it yourself, free, no sign-up.
Now, what actually matters when you’re choosing one of these for your own business, and it’s not the brand names or the marketing copy.
The first thing to understand is how the question actually gets from you, typing it into a chat box, to an answer showing up on the screen.
With both of these, you’re not just searching a database — you’re asking a real question, in plain English, and the system goes out and constructs an answer based on what’s actually available, in real time, every single time.
But there’s a big difference in how each one actually does that.
On the left, with Veo, think of it as a guided conversation.
When you ask something, it goes out and pulls together whatever relevant sources it can find — a FAQ page on your site, an article you’ve already published, a pre-written reply if there’s a common question it recognizes — and it structures that into an actual conversation, with a beginning, a middle, and an end, read out loud, the same way a real person might explain it.
On the right, with Gemini Pro plus the Omni add-on, think of it as a live search—it fires the moment you hit enter and pulls results in real time, instead of structuring them into a conversation.
How It Gets You Found
Now let’s talk about how these actually get your listings in front of buyers. Here’s the real math, quickly. Both tools work a state at a time — you pick the state, they build the map, you publish. On Google, the free version gives you one map per month, published on a schedule — that’s 12 maps a year, automatically, doing the heavy lifting of staying current for you. The version you pay for adds a team seat and removes the publishing cap, so the ceiling is whatever you can build in a month. On Gemini, the free version also gives you one map per month, published manually — you build it, you upload it, there’s no scheduled publishing. The paid version adds a team seat and automatic publishing, so it runs the same 12 maps a year as the capped Google version, without you touching it. Neither of these is the ceiling on what a skilled user can build — both offer uncapped subscriptions where the only limit is your own output — but those are the decisions the pricing makes for you, automatically, based on what you’ve signed up for.
What It Means For You
For you specifically — the agent trying to decide where to put your own resources — this means you can stop treating every new AI tool as if it’s the one you’ve been missing. You now have a repeatable way to look at any of them and understand exactly what it’s doing behind the language, and whether that’s actually the gap you have in your own business.
If you’re building your own brand, this is how you stop handing your content work over to a platform you don’t own. You become the one who can actually answer the question, in public, backed by sources, instead of asking Google and getting someone else’s answer read out loud.
And if you’re pushing for growth, this is how you stop buying attention you can’t actually convert. You walk away from the cheap click bait and put your marketing budget into the few channels you know, with certainty, are getting in front of the right people. You become the one who can answer the CEO question — what’s actually working, and by how much — instead of the one hoping something gets lucky.
What It Frees You Up To Do
This is the real math, and it’s why we walked away from both. Neither a rented AI tool nor a revenue-sharing one makes us more profitable, because the cost comes straight off the top of every single deal it touches — and that’s true even on a slow month when the tool sits idle. We didn’t build our business on renting bandwidth from someone else’s machine, and we didn’t get here by sharing our income with a platform that’s doing work we could give to an employee. Both of those models scale the same way — they get more expensive as you get busier, which is the exact opposite of how a growing business should be designed.
That’s really what this comes down to: ownership versus rentership. We didn’t swap from one AI tool to the other — we walked away from the rental entirely and put the same functionality on our own machine, running our own code, on our own schedule, for zero recurring cost. That’s what makes us more powerful than either of the rentals, and it’s the exact same move we teach to agents who are serious about building something they actually own.
If you want to run this on your own business instead of watching us do it on YouTube, here’s the honest version of what we just described. It’s not a software download — we’re not selling anything. It’s a free call. You book it right from the screen, and we’ll show you exactly how we run this on our own business, what it costs us in raw numbers, and whether it fits what you’re building. The link’s in the description, and our number’s right there too.
What To Do First
If you’re building an AI-powered real estate brand, here’s what you do not need to do first: you do not need to become a developer, and you do not need to build anything from scratch.
The move is to start from what already exists — two, specifically, and we’ve put both on screen. One is Google’s own AI video tool, called Veo. The other is a brand-new one, still in preview, called Gemini. Both do the same job: they take a short text question, in plain English, and they produce a video answer, on screen, live, in seconds.
Here’s the mistake to avoid right off the top. Do not treat these two as the same thing, or as two versions of the same idea. They are not. They work in completely different ways, they produce different kinds of answers, and the right choice depends entirely on what you’re actually asking.
So let’s open both windows and walk through the same question with each, side by side. We’re going to ask both the same thing: what are the school districts in a given county, and how are the schools performing. That’s a real question, a practical question, the exact kind of thing a buyer or a seller actually asks.
With Google Veo, you type that question into the chat window, and the on-screen avatar reads it back to you, then answers it, live, in a few seconds. The answer comes back as a short paragraph, spoken on screen, and it is pulled, live, from Google’s own sources — which means it is current, it is the same information you would get if you searched the same thing yourself, and it is the same for every user, every time, because it is pulled fresh each time someone asks.
Now, the big difference, and the reason this is the right choice for some things and not others. Because the answer comes from a general knowledge source, it is general. It gives you the names of the school districts and a broad, published description of how they’re performing — how many students they serve, what the student-teacher ratio is, whether the district is classified as ‘meeting standards’ or ‘needs improvement’ or something else, straight from the official classification.
But it does not give you anything deeper or more specific than that, and it should not.
Where to Start
If you want help building or evaluating a similar system for your own business, we offer a free 15-minute call — the link is in the description, along with a free starter guide you can download right now. No pitch. Just two agents who teach this stuff every day, sharing what we’ve learned.
Subscribe so you don’t miss the next live session — we build one of these, on screen, from scratch, every single month. We’ll see you in the next one.
Frequently Asked Questions
What’s the difference between Google Veo and Gemini Omni?
Google Veo works as a guided conversation that pulls together relevant sources like FAQ pages and articles, then structures them into a conversational answer read out loud. Gemini Pro with the Omni add-on works as a live search that fires in real time, physically loads your site in a browser tab, and walks through user flows the way a real visitor would, then delivers a written report with specific findings.
How does the basic search layer work differently from the AI layer?
The basic search works one way: one line, one question, one result, searching raw site code and handing back unfiltered matches. The AI layer sits on top of it, rephrasing unclear questions into searchable format, widening vague searches to include surrounding areas, and filtering results into a quick-read list—but it pulls from the exact same source, so there’s zero risk of getting false information.
Which tool should I choose for my real estate business?
The choice depends entirely on what you’re trying to do. Use the basic search for exact matches when you have a clear question. Use Google Veo if you need general knowledge answers structured as a guided conversation. Use Gemini Omni if you need to simulate how a real user would navigate your site and pull findings from multiple sources including local news and neighborhood guides.
Here’s how to set up a question-answering system that knows your market and runs on schedule in the background. Most agents get generic answers — ‘get more leads’ — even when the real question is something specific, like ‘why is my listing sitting.’ Today we’re showing you how we run that review automatically, on demand, with Claude Code.
If you’re a few years into this business — closing 6, 7, 8 deals a year, pushing for more lume, more consistency — you’ve probably already built or bought some kind of help: a follow-up sequence, a report, a posting schedule. The question now isn’t ‘can a chatbot do that,’ it’s ‘can a chatbot do that and get out of the way when I need to actually talk to a client.’
Both of these tools can handle the question without you ever touching it — you just tell the chatbot what the question’s actually asking, in plain English, one time, and from then on it’s a saved query you can run from the command line or pull up in the app, the same way you’d run a Google search you’d saved as a shortcut.
Where they start to differ is the ‘posting schedule’ job. With ChatGPT, you’d set that up by telling it ‘post X on Monday, Y on Wednesday, Z on Friday,’ and it would write that schedule down and follow it — but if you wanted to change what gets posted on a Tuesday, you’d have to go back and tell it that, in your own words, every single time. With Claude, you set up that same schedule once, and then you change it from the settings, the same way you’d change a recurring alarm on your phone. The instruction is the same whether you’re adding one extra day or switching the whole thing to a different rhythm — you’re not re-teaching the bot every time you want to adjust it. That’s the real difference between being a saved question and being a running system, and it’s the difference between a tool you use occasionally and a tool you actually depend on.
What It Actually Does
Here’s what both actually do when you ask them a question like that. A query like ‘what’s the market doing in my top 3 zip codes’ goes to Claude as a plain English question, and it comes back as a plain English answer — not a chart, not a link, but an actual reading of what’s happening, pulled from the sources it’s been told to trust.
On ChatGPT, the same question becomes a request for a specific format — ‘give me a market update for these 3 zip codes, formatted as a short paragraph’ — and it writes that paragraph. It doesn’t go out and collect data; it writes an answer on the spot, using whatever it’s already stored, and it’s the same every time you ask, because it doesn’t re-check the sources.
Now, both of them can sit on top of a structured data source — a spreadsheet, an API, a database — and if that source is set up right, both can read it and hand you an actual, current, formatted report instead of a written paragraph. The difference shows up when the question isn’t about a specific set of zip codes or a particular kind of report — it shows up when you ask something open-ended, like ‘what’s happening in my market this week.’
Claude will go back to the sources, re-check them, and give you a fresh reading, because the question is about the current picture, not a stored answer from last month. On ChatGPT, that same question comes back as the same written paragraph it’s already given you before, because the question doesn’t tell it to go re-check anything — it just tells it to give you the market update it already wrote.
How You Set It Up
Here’s the truth about both, straight from the source. On screen is Claude, built on the same code that runs Google’s own AI products. It costs $20 a month after a free trial. The free version has limits and gives priority to paying users. ChatGPT is the one that’s been in the news — it’s from OpenAI. It has a free version with unlimited chat, no trial, and no priority system. Both have free versions — Claude’s free tier has limits, ChatGPT’s has unlimited chat. Claude’s paid tier is $20/month. If you only use free, it costs nothing every month.
Here’s what actually matters. Both answer questions when you ask them — you open the chat window and type ‘what’s the status of the Smith listing,’ and it tells you, right there in the chat, the same way you’d ask an assistant. Neither runs fully automated in the background without you asking, but you can set up saved queries to rerun on a schedule. There’s no app to open, no dashboard to load, no daily routine to remember. You just ask, and it answers. That’s it.
Here’s the setting that actually makes them useful, and it’s the same for both. Think of it as a shared brain that works alongside your own. Every morning, you write down the 3 or 4 things you actually need to accomplish that day — a follow-up call, a report you need to pull, a posting you need to schedule. You don’t type them into either AI. You write them down, the old-fashioned way, on paper or in a note on your phone. Then, anytime you need help with any of those items, you ask the AI the same way you’d ask an assistant — ‘what’s the phone number for the Smiths again,’ or ‘what’s the headline I should use for today’s market update.’ It pulls the information and gives it to you, right there in the chat, the same way it always has. And here’s the part that actually saves time: if you’re in the middle of something and you realize you need a fact or a source you know you have somewhere, you don’t have to get up and go find it. You just ask the AI, the same way you’d ask an assistant standing next to you. ‘What was the median sale price in the last quarter.’
How It Gets You Found
People search for an agent by name, by location, and by the kind of help they need — ‘best agent near me,’ ‘how to sell my house,’ ‘what’s my home worth.’ Both tools can write a reply that answers a question like that, but only one hands you a search-ready page you can publish in 2 minutes, without a developer or a content team.
Claude Code gives you a fill-in-the-blank template you open in Google Docs, and the target page is the one that comes back when you Google the question — the same way our own published pages get found. ChatGPT gives you an answer you can paste into a blog post, but the post itself is not set up to be found. It does not come back when you Google the question that was asked.
If you want to run this on your own machine, the free starter skills are open-source on GitHub — just search for ‘Claude Code real estate skills’ and pick the one that matches what you need.
What It Means For You
If you’re pushing for lume, this points you to ChatGPT for the pipeline work — the calls, the follow-up, the checking. If your business runs on relationships and referrals, Claude is the one that protects your existing base while it’s working on the new leads. And if you’re building a brand, you now have two engines, running two different content directions, against one set of sources — split the questions between them and you get a content calendar that’s actually two conversations instead of one script read aloud.
Both of these are free trials, and the trial versions run the exact same way as the paid ones — the difference is the number of concurrent conversations, and that only matters if you’re trying to run more than 2 or 3 of these in a row at once. We are not affiliated with either company — we just run them in our own business every day.
What It Frees You Up To Do
The follow-up calls, the reports, the posting schedule — those are the same jobs you’d give a paid assistant. The fourth job, building a knowledge graph, is the one that turns into a repeatable system, because it’s like the owner telling you, over and over, what matters in their property, without you ever having to ask. That’s what makes you more efficient than a chat window full of one-off questions, typed out by hand, each time you start working with a new client. You’re building something you can point to, instead of renting someone else’s platform by the message.
What To Do First
If you’re watching this, you’ve probably already asked something like this to one of them: ‘Claude, what’s a good follow-up email to a buyer who went quiet after the third showing?’ Or: ‘ChatGPT, how do I respond to a client who’s frustrated that the market isn’t improving?’ Both will answer. But one of them will actually help you close more deals, because the difference is subtle — and most agents never notice it.
Here’s what you do first, before you ask anything else. Open a new Google Doc and write down every single question you’ve asked Claude Code or ChatGPT about your business in the last 30 days. Every question, big or small, whether it’s about a specific deal or just a quick confirmation. Do not open either AI and ask it to give you the list — write it down by hand, or type it out yourself. That way you’re actually looking at every question you’ve asked, not just the ones you can remember.
Once you have the list, number each question. Then go through it and add a column — mark each one as either ‘needs a source’ or ‘doesn’t need a source’. A question needs a source if the answer actually matters to a client, or if it’s something you’d normally look up yourself instead of asking an AI. ‘What’s a good email to a quiet buyer’ is the first kind — the answer matters to the client, and there’s actually a right answer that applies to your market. ‘Is it normal for a seller to ask about my commission’ is the second kind — that’s not a normal question for an AI to answer, it’s the kind of thing you’d ask a mentor or a broker.
Now here’s the part most agents miss. Open ChatGPT and ask it the same question you just asked Claude — ‘what’s a good follow-up email to a buyer who went quiet after the third showing’. Then do the same thing with Claude. Write down both answers. Almost always, the answer from the one you’ve been talking to for 30 days is a little sharper — it knows your context, it knows your market, it’s pulled from the conversations you’ve already had. The answer from the other one is usually a bit broader, a bit more generic, because it doesn’t have that history yet.
Where to Start
If you want to run this on your own business, the free starter skills are open-source on GitHub. The full automatic version is one of our private partner options. You can book a free 15-minute call with us or reach us directly.
We’re Al and Victoria Pinder, husband-and-wife ICON agents at eXp Realty — and we run this in our own business every single day. We’ll tell you honestly what fits your business and what doesn’t.
Frequently Asked Questions
What’s the main difference between Claude and ChatGPT for running scheduled market reports?
Claude lets you set up a posting schedule once and adjust it from settings like a recurring alarm, while ChatGPT requires you to re-teach it the instructions every time you want to change what gets posted. This makes Claude better for running a system you actually depend on, rather than a tool you use occasionally.
How does Claude handle open-ended market questions differently than ChatGPT?
When you ask Claude something like “what’s happening in my market this week,” it goes back to the sources and gives you a fresh reading. ChatGPT gives you the same written paragraph it already wrote before, because it doesn’t re-check the sources unless the question specifically tells it to.
Which tool helps you build searchable content that gets found on Google?
Claude Code gives you a fill-in-the-blank template that creates a search-ready page you can publish in 2 minutes, the same way pages get found when you Google the question. ChatGPT gives you an answer you can paste into a blog post, but the post itself isn’t set up to be found by Google.
You’re building your real estate brand on YouTube, and you know the single biggest challenge isn’t making the videos — it’s remembering to actually post them. A week goes by, you’re busy with clients and showings, and the next thing you know there’s a video sitting on your desktop doing absolutely nothing. It never occurred to anyone to check if you’d hit ‘upload’ on it, because from the outside it just looks like you didn’t feel like posting that week. That’s not a content problem, and it’s not a lack of motivation. It’s the five minutes between finishing the video and actually hitting upload — and those five minutes are the only thing standing between consistency and silence.
If your business runs on relationships — repeat clients, referrals, people who already trust you — you know your biggest challenge isn’t getting found by a stranger online. It’s staying top of mind with the people who already know you, without becoming that annoying friend who only talks about real estate. You want to be the one they think of when something changes, without ever feeling salesy. That’s exactly what we’re building this week.
What It Actually Does
Here’s the part that actually matters if you’re not technical: none of that matters to you. You don’t need to know how to code, or how any of this actually works under the hood. You just tell Claude what you want — ‘post this video on my YouTube channel and my Facebook page, at 8 AM, with this message’ — and it does the rest. It writes the automation, sets up the schedule, and tests it. Then, every morning at 8, the video goes out on both platforms, with the message you gave it, without you touching anything.
If you’re building a brand, this is what lets you stay consistent without adding hours to your day. And if you run on referrals, this is how you finally start getting found online by the people who’ve never met you.
How You Set It Up
Here’s the actual setup, because we do these and you can watch it yourself if you want. We demonstrate this workflow in our Prosperity Agent training materials, and the free starter skills are in the description word for word. The version we run on Claude Code is the one with the extra features — the full source is in the description, and you can clone it and run it on your own machine if you want.
The free starter skills just run on the command line. You open a terminal, type ‘claude code run youtube-autopost’ and it runs the whole thing — finds the new video, writes the post, schedules it, and tells you what it did. You don’t need to be a developer to run it, and you don’t need to stay at the keyboard while it runs. You can close the terminal and come back later — the scheduled post will still be in the queue ready to go out at the right time.
The first thing it does is check whether there’s a new video ready to post. It goes to YouTube and searches for your channel using the channel name you gave it when you set it up — and yes, you can run this on more than one channel, just give it the right one for the day. It looks at the upload date of the most recent video and checks whether that date is today. If it is, it means there’s a new video ready to go out. If it’s not, it doesn’t schedule anything — no sense posting last week’s video this week, right? That’s it for the free starter skills. Two commands, runs itself, and you’re done.
The version we run on Claude Code adds a few more features on top of that, and that’s really what this is about — showing you what Claude Code can build that the free skills can’t, and why that matters if you’re trying to run this on your own business instead of watching us do it on YouTube.
First, the free skills only schedule one post at a time, and they schedule it for tomorrow’s morning slot. The Claude Code version watches your videos the same way, but it schedules the whole rotation for the week in one pass — tomorrow’s post goes out first thing in the morning, the next one goes out the following evening, the next one the next morning, and so on. So even if you walk away after that first run and never touch it again, it keeps itself running for the whole week without you.
How It Gets You Found
Here’s the part that actually matters if you’re not a tech person, because neither are we.
If you want to skip the setup entirely, we also offer a done-for-you service. You fill out a form in plain English, tell us what you want the video to be about — a neighborhood spotlight works well — and we handle the rest: build the automation, upload the video, schedule the posts. Every Sunday night the next week’s content is queued up waiting for your approval.
If you’re pushing for pipeline, leads, and consistency — the automation removes that friction for you. You stop being the bottleneck, the content gets made, the pipeline keeps filling, and you get your time back for the actual relationships.
What It Means For You
If you’re building a brand on video, this means you can stop letting a single platform dictate your schedule — one system serves both feeds, and you pick which one gets the priority. If you’re stretching a marketing budget, it means one piece of content works three places instead of one, without adding hours of extra work. Either way, it’s the same few minutes of setup, running on a timer, while you get back to the camera.
What It Frees You Up To Do
Every week this runs is another week of consistency. That adds up. The pipeline notices.
That’s real bandwidth back — bandwidth you can put into the next client call, the next follow-up, the next relationship, instead of into a posting grid.
If you’re pushing for pipeline, this is how you serve the pipeline you’ve already built, without adding hours to your day.
And if you’re building a brand, this is how you get found by Google and by the AI tools, consistently, without becoming a full-time content creator.
You own your content, you own your audience, and you built the machine that feeds it — without adding hours to your week.
What To Do First
If you’re pushing for pipeline, the instinct here is usually to build out a content factory: more posts, more videos, more pages, faster. And that instinct is not wrong — quantity has value, especially when you’re starting from zero.
The problem is that most of us do not have the team or the hours to run a content factory by hand. We are one person trying to serve clients, follow up on calls, and prepare for every showing and every closing.
So the first thing we tell anyone who asks us this is: do not build your own content factory. That’s exactly what Prosperity Agent does — it takes your videos and your blog posts and publishes them on a schedule across all of your social profiles — Facebook, Instagram, LinkedIn, Pinterest, even your website — while you sleep, or while you’re at a showing.
Every morning, your content is sitting in all of the right places, published on schedule, and you didn’t touch it. That is not a shortcut for one post — that is a machine that runs in the background, on a schedule, doing the work that used to eat an entire afternoon.
Now, there is a preview mode. Before anything goes out, you see it exactly as it will look — the headline, the image, the text — and you can approve it or edit it right there. Nothing goes live until you’ve seen it and approved it.
The approval email arrives while you’re making your morning coffee. By the time you sit down, you have a real, dated, publishable piece of content that you did not personally spend an hour writing or scheduling.
That is the whole point. You are not spending your time building a content system — you are hiring one, and that system runs on a schedule while you get your morning back.
If you want to see exactly how we set this up — the before-and-after, the exact settings, the whole thing running live — that is what the free starter skills are for. You can open the skills and look at the ‘Content Factory’ workflow yourself, no cost, no sign-up. It runs on your own computer, on your own schedule, and you watch it work before you ever commit to anything.
Where to Start
If you want to run this on your business, here’s your next step: book a free 15-minute call with Al and Victoria Pinder to see what fits for your situation., and we’ll look at your business and tell you honestly what fits and what doesn’t.
The link to book it is right in the description below, and our number’s there too.
And if you’d rather watch this on your own first, the free starter skills and the skill setup guide are also in the description — open source, no charge, they run on your own machine, and you are welcome to them whether we ever speak or not.
What’s the biggest challenge with posting real estate videos consistently?
The main problem isn’t making the videos — it’s remembering to actually upload them. Between finishing a video and hitting upload, you get busy with clients and showings, and the video ends up sitting on your desktop doing nothing. Those five minutes between completion and posting are what stands between consistency and silence.
How does the automation actually work if I’m not technical?
You simply tell Claude what you want — like “post this video on YouTube and Facebook at 8 AM with this message” — and it handles the rest. It writes the automation, sets up the schedule, and tests it, then every morning at the scheduled time, your content goes out on both platforms without you touching anything.
Can I approve content before it goes live?
Yes, there’s a preview mode that shows you exactly how the post will look — headline, image, and text — before anything goes out. You get an approval email while making your morning coffee, and nothing publishes until you’ve seen it and approved it.
You’re building your real estate business on relationships — repeat clients, referrals, the people who already know you. And you’re pouring your own hours into keeping up with every one of them, every property, every thread. You know the sharp turn the day takes the moment a client asks something you didn’t see coming, or a detail slips through because there just weren’t enough hours in the last one. That’s not a sign you’re doing it wrong. That’s the natural ceiling of ‘me plus a spreadsheet’.
This one’s for the agent closing somewhere around 6 to 8 deals a year, looking to scale that number without adding hours that don’t exist — and for the agent who built their business on relationships and trust, now figuring out how to hold onto that while everything goes digital. If you’re a few years in, pushing for more volume, and you’ve already got the CRM, this is the next real question: do you build or buy the AI layer on top of it, and which one actually gets the queries right?
What It Actually Does
With ChatGPT, the answer is mixed. It’s great at the basics: a quick fact, a simple calculation, a standard reply. It can sound human in a back-and-forth. But give it something complex — a genuine objection, a true confusion, a real question about a deal — and it starts to guess. It doesn’t have context for a specific transaction, it can’t pull a real record, it can’t check an actual file, it can only suggest an answer it’s been trained to give, and sometimes that suggestion is wrong.
Now Claude. Same test, different result. It doesn’t guess, because it can actually check. It can open a real file, pull an actual record, look at the actual language used, and answer from what’s actually there — not from a memory of something it was once fed. That makes a real difference when the question isn’t a hypothetical. It’s the difference between a chat window that spouts generalities and a tool that actually helps close a deal, because it’s actually looking at the deal in front of it.
Here’s an example, because examples are what matter. Both systems get the same question, typed the same way, no difference in how it’s asked. Which one is it? The difference isn’t that one’s smarter than the other. It’s that one was built for conversation and the other was built for a specific job, and you can see the difference in every real question you actually ask them.
How You Set It Up
Here’s the setup, and we’re keeping it simple on purpose, because the same principles apply to bigger, more complex setups.
We took two Windows laptops, ran the same specs on both — 16 gigs of RAM, i7 processor, SSD drive — so it’s the software doing the work, not the machine it’s running on.
On one, we installed Claude Code with the free starter skills — that’s the version we’re partners with, the one you can try yourself — and we added the real estate skill pack on top of it, which is $49 a month, the same way a couple of the agents who’ve already signed on have done.
On the other, we set up the free version of ChatGPT, which runs in a browser window and costs nothing.
Then we gave both the same job, in writing, the same way you’d brief an assistant you’d hired.
We didn’t teach either one how to be an agent — we wrote out the kind of task you’d actually hand to an assistant, the way you might write a note to your own team: what needs doing, by when, what the client told us, what the rules are for this one.
Three things we wanted to watch, specifically, because these are where the moves actually matter.
First, does it understand the instruction the first time, or does it ask for a repeat or a clarification?
Second, does it stick to the rules you gave it, or does it wander off and generate something you never asked for?
And third, once you tell it to stop working on one thing and start on something else, does it pick up the new task and put the old one away, or does it keep re-litigating the first job instead of moving on?
Those three are what actually matter in a business setting — understanding the ask, staying on topic, and responding to a change in priority.
Now, the results.
On the Claude Code machine, the run looked like this.
It read the brief, nodded, and got to work.
It checked our own website first, which is exactly what you tell a new assistant to do — look for the answer in the sources we already pay for before you go outside.
Then it went to Realtor.com, to Kelly Blue Book, and to the county’s own online record system, pulling the right property address each time, not a generic town page.
How It Gets You Found
Here’s the real test, outside of us typing at it: how each one handles a brand-new, specific question that’s actually asked of an agent, every single day, with no lead-in, no setup — just the question, typed into a chat window, cold.
We ran the same three questions on both, with the same account, the same day, logging every response.
Question one: ‘What’s the market doing in our town right now — is it getting busier or quieter?’
Claude gave a clear, local read, sourced from a real, current news outlet, and it was written in plain English — not a headline dumped on us, actually parsed into what it means for the market.
ChatGPT gave us a generic, global take on ‘real estate market trends,’ with zero mention of the town, and zero sign of anything current — it pulled from a old, general guide, not a real, recent source.
Question two: ‘Can you run a quick check on a house that just came on — is the listing accurate, or is something off?’
Claude checked the listing against the actual public deed record, found a real discrepancy — the square footage listed was off by 400 — and told us exactly what didn’t match, in plain terms, without us asking for that specific comparison.
ChatGPT gave us a generic ‘here’s how you verify a listing’ guide, with no actual check run, no specific house mentioned, and no real answer to the question asked — it just described, in the abstract, how you might check something, without actually checking anything.
Question three: ‘I’m thinking of listing my house soon — what’s a reasonable time to be on the market right now?’
Claude looked at the current season, checked how long similar homes were taking, and gave us a genuine, specific range — ‘Right now, 4 to 6 weeks,’ with why: fewer school-season moves and holidays coming up.
ChatGPT gave us a generic ‘it varies’ answer, with no reference to the time of year, no reference to what’s actually happening in the market now, and no specific number or range — just a vague, fill-in-your-own-numbers answer that didn’t actually answer the question.
Now, one real takeaway: these aren’t the same tool doing the same job. They’re built for different kinds of work.
What It Means For You
For you specifically — the agent trying to choose between these two — it means you can stop thinking of a chatbot as one generic tool and start thinking of it as two tools, built for two different jobs. If your question is about a client or a deal currently in your pipeline, Claude is probably the better choice right now, because it’s actually looking at your own data while you ask it. If your question is about getting smarter — staying on top of a changing market, a new town you’re trying to break into, or a niche you want to specialize in — then ChatGPT is the one more likely to give you a useful, panoramic answer, because it’s been exposed to so much more source material than Claude has so far.
Both of these are free to try right now — Claude offers a free, no-credit-card-needed starter tier, and ChatGPT has a free version you can access just by signing up.
What It Frees You Up To Do
The real value shows up in what it frees you up to do.
You’re not sitting down to write a blog post at 9 PM after the kids are in bed — you’re approving a draft the AI wrote, maybe with one fact checked and one section rewritten, and then you’re back to whatever needs you next.
That’s real hours back, not a trick or a feature.
On a client call, instead of trying to remember the 4 or 5 different programs a property might qualify for, you’re showing them a short, correct list the AI pulled together in 90 seconds, sourced from the actual program guidelines, while they’re still on the line.
You look prepared because you are prepared — not because you spent an hour digging through a website beforehand.
That’s what closes deals and builds a reputation: being sharp in the moment, every time, because the heavy lifting was done before you even picked up the phone.
What To Do First
If you’re watching this, you’ve probably already asked something like this to ChatGPT or Claude Code: ‘Can you write a real estate listing?’ ‘Can you draft an FAQ for my website?’ ‘Can you review my contract language?’
Both tools will answer, and both answers will look right. The real question isn’t can they write something — it’s whether you trust it enough to send to a client. Should you check every answer against a real attorney, or run it by something built specifically for real estate first?
If you’re pushing for volume, that’s real hours back. If you’re building a brand, it’s the gap between consistent, correct content and whatever a chatbot happened to type. If you’re protecting your license, it’s the difference between a suggestion you vetted and an instruction you followed because a chatbot sounded authoritative.
None of that means you should never use these tools. Both have real, useful places in a productive agent’s workflow. The mistake is treating them as the last word instead of the first draft, and assuming anything they generate is ready for a client or a public page without a second, informed look. That’s true whether you’re asking a chatbot or an AI code interpreter — Claude is just as capable of generating something a lawyer would push back on as ChatGPT is.
The fix is not about any one tool or platform. It’s about how you actually use AI in your business, instead of just asking it a question and taking whatever it gives you.
Here’s the right order, the one we actually recommend to the agents who pay us for this stuff.
First, state the task as a question, clearly and specifically. ‘Write a listing description for 123 Main Street’ is plenty clear. ‘What do people ask when they’re buying a house’ is too vague — you’ll get a generic list that may or may not match what your actual clients ask.
Second, run that question against Claude Code first. Why? Because the version of Claude you get with our link is already set up with our own real estate knowledge graph running behind it. That means it’s already been primed with real estate terms, real estate scenarios, and real estate sources, before you’ve ever asked it anything. When you run that question through it, you’re querying something that’s already built for your business, not typing a question into a blank chat window and hoping it guesses right.
Where to start
If you want to run this on your own business — not just watch us do it on YouTube — here’s your next step. We offer a free 15-minute call. It’s just us, and we’ll look at one of your listings and show you how we’d apply this to it. Subscribe so you don’t miss the next live session — we build these on screen every month, and the Q&A is open to everyone who’s watching live.
Frequently Asked Questions
What’s the main difference between Claude and ChatGPT for real estate agents?
Claude can actually check your real files, records, and data to answer specific questions about deals and transactions, while ChatGPT generates answers based on what it’s been trained on, which means it guesses on complex questions and can’t pull actual property records or client information. Claude looks at what’s actually there; ChatGPT suggests what it thinks should be there.
Which AI tool should I use for questions about my current deals and clients?
Claude is the better choice for questions about clients or deals in your pipeline because it can access your actual data and provide specific answers about real transactions. ChatGPT is more useful for broader market questions and learning about new topics, since it’s been exposed to more source material overall.
How much does Claude with the real estate skill pack cost?
Claude Code is free to start with, and adding the real estate skill pack costs $49 a month. ChatGPT also has a free version available, so you can try both without spending money before deciding which works better for your workflow.
