Plain Answers · Real Estate
Can realtors use AI with client information?
Listing copy, yes. Their bottom line, slow down.
For listing descriptions, market summaries, and social posts — yes, freely. For anything a client told you in confidence, it depends on the tool, the plan, and your broker's policy, because no AI product is "compliant" on its own. The thing you're protecting isn't a file format. It's the fact that your buyer will go higher if pushed, and by how much.
Rules that may apply to you
Depending on what you do and who you serve, you may be subject to rules such as the ones below. We link the official sources and leave the interpretation to your broker, your brokerage's counsel, or your association's legal resources. This is a starting point, not a complete list, and nothing on this page says what these rules require of you.
Why this is a bigger deal in real estate than it looks
Real estate isn't usually listed among the "regulated data" professions, so agents tend to assume the privacy conversation is for doctors and lawyers. But look at what's in an ordinary transaction file: pre-approval letters, bank statements, tax returns, a divorce decree explaining why the house is being sold, a relocation deadline, and the number your client will really accept.
California agents may be subject to fiduciary and confidentiality duties to their clients. The statutory agency disclosure form — the one agents hand to clients, linked above — is the plain-language place to see them described. It's two minutes of reading, and worth doing with AI in mind.
We won't tell you how a regulator or a court would apply those duties to an AI tool. We'd only point out that the information a client most expects you to guard — financial position, motivation, price — is precisely what agents are most tempted to paste into a chatbot when they ask it to "help me write a counteroffer."
Brokers: this lands on your desk
Brokers may be subject to supervision responsibilities for their salespersons under the Business and Professions Code (section 10177 is linked above). How those apply to AI use is for your counsel. The practical point is simple: if thirty agents are each making up their own AI rules, the brokerage doesn't have a policy. It has thirty.
Larger brokerages may also be subject to the California Consumer Privacy Act. The Attorney General's page, linked above, describes who the law covers; your counsel can tell you whether that's you. If it is, the AI vendors that touch consumer information are part of that conversation.
A working split: what goes where
- Fine in most tools: listing descriptions from public facts, neighborhood write-ups, open-house posts, explaining escrow to a first-time buyer in plain English, summarizing a public market report.
- Only in a tool your brokerage has vetted: drafting emails that mention a client by name, summarizing an inspection report, organizing a transaction timeline.
- Keep out of consumer chatbots entirely: pre-approvals, bank and tax documents, IDs, anything about why a client is moving, and any number that reveals their negotiating position.
What to ask about any AI tool — including the ones built into your CRM
Much of the AI agents use now arrives inside software they already have: the CRM, the transaction platform, the email client. The questions are the same.
- Whose AI is it, really? Many tools pass your text to a larger AI provider. Ask which one, and under what agreement.
- Is what I type used to train or improve models?
- How long is it kept, and what happens when I delete it or leave the brokerage?
- Can the vendor's staff read it?
- Is this a personal account or a brokerage account? If an agent leaves, who holds two years of client conversations?
That last one is easy to overlook. Our walk-through of what happens to data in an AI chatbot shows how to check the rest in about ten minutes.
What we check when we deploy for a brokerage
- Agent-level separation. In a shared office system, one agent's client files aren't visible to another's AI queries. That matters most when both sides of a deal are in-house.
- Documents processed locally. Pre-approvals, disclosures, and inspection reports are read and summarized on hardware the brokerage owns, not uploaded to a public service.
- Accounts belong to the brokerage. When an agent moves on, the history stays, and their access ends the same day.
- No outbound path from the AI server, tested with the uplink unplugged.
- A one-page policy agents will actually read, built around the three bins above.
What it costs
A solo agent doesn't need a server. Our desktop app, Archivist, runs entirely on one Windows laptop and keeps every document local; the base version is free. For an office, a single-GPU in-house server runs $1,500 to $7,500 once, with no per-agent fee — which matters when your headcount changes every quarter.
None of it replaces your broker's policy or your own judgment about a client's confidence. For the general version of this question, see Is ChatGPT safe for business documents?