Most articles about AI in this industry are either a long list of tools or a prediction about the future. Neither one helps much when you're trying to decide what to do with your business this week.
This guide is written for wholesalers, flippers, and small landlords. It covers what AI agents in real estate actually do, which tasks they handle well, how to connect them to real property data, and where the legal risks sit. That last part gets the most space, because it's the part that can cost you real money.
AI Agent vs. Chatbot: What's the Difference?
First, a quick note on wording. In real estate, an "agent" is usually a licensed person. In this article, "AI agent" means software, and "licensed agent" means the person.
A chatbot, sometimes called an AI assistant, is a conversation tool. You type a question, and it gives you one answer. It can't do anything outside that chat window.
An AI agent works toward a goal across several steps with limited supervision. It can search the web, look up records, connect to other software, and take action based on what it finds. What this allows you to do is hand off a whole sequence of work instead of a single question.
- A chatbot writes a follow-up text for you to send.
- An AI agent pulls the lead from your CRM, checks the number against the Do Not Call list, writes the text, sends it, logs the reply, and books the callback.
Keep that in mind when you read software marketing. A lot of what's sold as an "AI agent" is really chatbot work with a nicer screen. There's nothing wrong with that, but it helps to know which one you're paying for.
What AI Agents Handle Well (and Where They Fall Short)
The table below lists common investor tasks and labels each one in one of four ways. Here's what each label means:
- Assistant: The AI works like the chatbot described above. It reads, writes, or analyzes something you give it, and then a person reviews the result and decides what to do next.
- Agent: The AI completes several steps on its own, such as replying to a lead, logging the conversation, and booking a call, with little or no human review in between.
- Either: The task can be set up both ways. You can have the AI draft the work for you to approve, or let it handle the steps on its own.
- Don't automate: The decision carries enough legal risk that a person should make it every time.
The last column shows the main thing to watch out for with each task.
| Task | Agent or Assistant | What It Does Well | Watch Out For |
|---|---|---|---|
| Contract and document review | Assistant | Reads long documents and flags unusual terms | It can be confidently wrong. Use it to find questions for your attorney. |
| Comps and ARV | Either | Pulls recent sales and suggests a value range | Least reliable on distressed or unusual homes |
| Marketing copy | Assistant | Drafts letters, ads, and posts quickly | Fair housing risk in phrases like "family friendly" |
| Speed to lead | Agent | Responds to new leads in seconds and books calls | Fast replies to a bad list still don't create deals |
| Text follow-up | Either | Sends sequences and sorts replies | Missing a plain-language opt-out creates a violation |
| AI voice calls | Agent | Places or answers calls with a synthetic voice | Highest legal risk on this list. Don't start here. |
| Tenant screening | Don't automate | Organizes application documents | You stay responsible for fair housing, even if software decides |
Two good places to start
Document review. This is one of the strongest uses today, and it carries very little legal risk because nothing gets sent to anyone. Give the AI a lease, title commitment, or purchase agreement and ask what's unusual, what falls on the buyer, and what's missing. It will surface good questions. It won't replace your attorney, and Stanford researchers found that even legal AI tools give false answers in roughly one out of six test questions or more.
Deal analysis spreadsheets. Have the AI build the spreadsheet, then check the formulas once. Don't ask it to do the math deal by deal. A formula gives the same answer every time and is easy to audit, while AI arithmetic is neither. Our walkthrough on how to analyze real estate deals covers what belongs in that model.
Connect Your AI to Live Property Data
The biggest change over the past two years is how AI tools get their information.
Until recently, you had to copy data into a chatbot and hope it read it correctly. Now, a standard called the Model Context Protocol (MCP) gives AI tools a common way to connect to outside data sources. Once a tool is connected, it can look up live records on its own instead of guessing from what it learned in training.
This is the practical line between an assistant and an agent. If the AI is recalling information, treat its local numbers with caution. If it's querying a live data source, you can work with the results much more confidently.
DealMachine works this way inside the AI tools many investors already use. With the DealMachine connection for ChatGPT or the connection for Claude, you can ask in plain English for things like:
- Vacant, high-equity properties in a specific zip code
- Owner details, equity, and mortgage information before a call
- Comparable sales for a property you're evaluating
- A filtered prospect list you can save and work from
Setup takes a few minutes. You sign in to DealMachine, approve the workspace you want to use, and then choose DealMachine inside your chat. Your normal account permissions still apply, so you can check your usage before you export anything.
This short video walks through the process from start to finish.
The Compliance Rules That Matter
What follows is general information, not legal advice. Please talk to a licensed attorney before you run any outbound campaign.
AI voices count as prerecorded voices
In February 2024, the FCC ruled that AI-generated voices are "artificial" voices under the Telephone Consumer Protection Act (TCPA). That means every existing consent rule applies. Damages run $500 to $1,500 per call, with no cap.
To use AI voice calling legally for marketing, you generally need:
- Prior express written consent that specifically covers an artificial voice
- Calls only between 8 a.m. and 9 p.m. in the recipient's time zone
- Do Not Call scrubbing and immediate honoring of opt-outs
There's an important difference between AI helping a person make calls and AI being the caller. When a person dials and AI transcribes the call, updates the CRM, and drafts the follow-up, you avoid the artificial voice problem and still get most of the time savings. This conversation with an investor using AI cold calls shows what the day-to-day side looks like.
Opt-outs don't have to use your keywords
Since April 2025, FCC rules let people revoke consent in any reasonable way, and you have 10 business days to honor it. "Stop" and "unsubscribe" count, but so does "please stop contacting me about my mother's house."
This is where automated reply sorting can fail. Test your system against real messages from sellers, not just a keyword list.
Some states require you to disclose AI
California has banned using a bot to hide its artificial nature in sales conversations since 2019, and the disclosure has to appear in the conversation itself. Utah requires a clear answer whenever someone asks if they're talking to a human or AI. If a seller asks your system "Am I talking to a real person?", it needs to answer honestly.
Texas also expanded its telemarketing law to cover text messages starting September 1, 2025. We cover the details in our breakdown of Texas SB 140.
Fair housing responsibility stays with you
Fair housing liability can come from discriminatory results, not only intent. Using a vendor's AI doesn't move that responsibility off your plate. Three areas deserve extra care:
- Marketing copy. Review every AI-written ad for phrases like "safe area" or "great for young professionals."
- Ad targeting. Avoid narrowing audiences in ways that stand in for a protected class.
- Tenant screening. Let AI organize documents, and let a person make the decision using written, consistent criteria.
Keep seller data out of personal chatbots
Many consumer AI plans may use your conversations for training unless you turn that setting off. Business plans usually don't by default. Before you paste a seller's name, phone number, or financial situation into a chat, turn off training in your settings or use a business plan.
Does AI Actually Pay Off?
Adoption is high, but results vary. In the 2025 NAR technology survey, about 68% of Realtors had used AI tools. Even so, 46% reported no noticeable impact on their business, and only 17% reported a significantly positive one.
The payoff comes from specific, well-defined tasks rather than from buying a platform and expecting it to change everything.
Where to Start
Here's a sensible order, starting with the lowest cost and lowest risk:
- Pick one paid AI tool. A general plan at around $20 a month covers most investors. One subscription is usually enough.
- Turn off training on your data. Do this before you paste in anything about a seller.
- Connect it to real property data. Live records make the answers far more useful than general knowledge.
- Use AI to draft, and have a person send. This keeps most of the benefit while avoiding the riskiest rules.
- Add AI voice only with an attorney. You'll need written consent, disclosure, Do Not Call scrubbing, and any required state registrations.
From here, pick one task from the table above, use it for a month, and measure whether it saved you time. If you'd like to see what's included at each level, you can review DealMachine's current plans and decide what makes sense for you.
Frequently Asked Questions
Is it legal to use AI to cold call sellers?
Only with the right consent. The FCC confirmed in 2024 that AI-generated voices count as artificial voices under the TCPA, so marketing calls need prior express written consent that covers an artificial voice. Damages run $500 to $1,500 per call. A call where a person dials and AI only takes notes and drafts follow-up avoids this issue, which is why most investors should start there.
What's the difference between an AI agent and an AI chatbot?
A chatbot answers one prompt at a time and can't act outside the conversation. An AI agent works toward a goal across several steps and can look up records, connect to other software, and take action. For example, a chatbot drafts a text, while an agent checks the Do Not Call list, sends the text, and logs the reply.
What are the best AI tools for real estate investors?
Most investors only need one general AI tool, such as ChatGPT or Claude, connected to a reliable property data source. The general tool handles writing, analysis, and document review. The data connection gives it accurate property, owner, and comp information so it isn't guessing.
How accurate is AI for estimating ARV?
It's useful for screening but not for setting an offer. Automated valuations are least reliable on distressed, damaged, or unusual properties, which are often the ones investors buy. Use AI to decide which properties deserve a closer look, then confirm value with local comps and a walkthrough.
Do I have to tell people I'm using AI?
In some states, yes. California requires disclosure when a bot is used in sales conversations, and Utah requires an honest answer when someone asks whether they're talking to AI. TCPA consent for AI voice calls must also mention the artificial voice. Rules vary by state, so check with an attorney for the states where you market.