eXp AI Rules for Client Data, Marketing and Follow-Up
AI can help organize and draft work, but a useful workflow must account for client information, factual accuracy, disclosures, and human review. This guide applies the public eXp policy to practical marketing and follow-up situations without assuming that every tool connection is approved.
By Chris SanfordSources checked October 27, 20265 min read
The direct answer
eXp’s published U.S. policy requires agents to protect client information, verify AI-assisted facts, and follow specified disclosure and marketing rules. Do not upload private client files into public generative-AI tools or assume a paid subscription makes every use acceptable. Start with a narrow task, use only permitted information and tools, review the output yourself, and obtain the current policy language for the intended use.
Key takeaways
- Tool access does not establish permission to disclose client information.
- Use a controlled drafting and verification process before publication or sending.
- Confirm current company and state disclosure requirements for the actual output.
Read the policy as a workflow requirement
The August 2026 U.S. policy, Data Privacy and AI section addresses data handling, human verification, disclosures, and altered imagery. Read the current version before launching a workflow. A product demonstration or successful integration does not answer whether a specific input, output, or communication is permitted.
Define the task first. “Help me outline a generic seller checklist” has a different information requirement from “summarize this client’s signed contract and financial documents.” The practical discipline is to minimize the data involved and understand every place it will go. If you cannot explain the workflow clearly, simplify it before adding real client information.
Classify the proposed task before choosing a tool
| Task | Safer planning approach | Question to resolve |
|---|---|---|
| Generic educational draft | Use fictional or nonpersonal inputs | What facts need verification? |
| Client-specific analysis | Use only an authorized process | Is this data and tool use permitted? |
| Automated follow-up | Review identity, consent, and escalation | What disclosure and human review apply? |
| Edited listing image | Preserve truthful property representation | What labels and original-image rules apply? |
This table is an original planning aid, not a substitute for the policy. It helps identify the right questions before information is entered. Treat tool settings, data storage, access, and retention as part of the decision. A convenient feature should not become a reason to skip understanding the underlying information flow.
Use a drafting process that preserves factual control
For a hypothetical property-description task, prepare a verified fact sheet yourself and remove unnecessary personal information. Ask for a draft based only on the supplied facts. Then compare every factual statement with the source record. Delete invented details, unsupported neighborhood claims, and language that implies knowledge you do not have.
Keep a short verification checklist for dimensions, amenities, condition, dates, and any claim likely to influence a buyer. Do not treat fluent wording as evidence. If the draft adds a feature that was never supplied, correct the draft and examine whether similar additions appear elsewhere. The agent remains responsible for the final communication and should be able to explain its factual basis.
Keep automated follow-up accountable
Before automating messages, identify who receives them, why the communication is appropriate, what information the system uses, and when a human takes over. Test with controlled records. Review the content for false personalization, inaccurate status claims, and promises that your actual service cannot meet.
The company policy includes prescribed disclosure obligations; obtain the exact current wording and placement instructions rather than improvising a substitute. For commercial email, the FTC’s business guidance also addresses accurate identity and opt-out requirements. Different channels can have different obligations. A working automation is not proof that its communication is appropriate or compliant.
Handle images and sensitive information deliberately
A useful image workflow begins with the truthful representation you need to preserve. If you create a furnished illustration, retain the original and document what changed. Review the current brokerage, state, and MLS rules for labeling and presentation. Do not allow attractive output to conceal an issue a buyer needs to understand.
For sensitive information, stop before uploading if authorization is unclear. Ask about the approved process using a description of the data category rather than attaching the private file to a general support message. If an unintended disclosure occurs, preserve the relevant facts and use the current incident-reporting route promptly. The contact-data guide explains why access to information and permission to reuse it are separate questions.
Create a short approval and review record
- Name the task and the business reason for using AI.
- Identify the permitted inputs, tool, account settings, and access.
- Check the current company and state requirements.
- Test with fictional or appropriately limited data.
- Verify facts, representation, and required disclosures before use.
- Record the human reviewer and revisit the workflow when the tool or policy changes.
Use this process for a small number of valuable tasks rather than automating everything at once. The AI of Choice overview describes the separate platform initiative; a connection should still be evaluated within the applicable data and communication rules. Start with a workflow you can supervise, then expand only when you understand its behavior and limits.
Common follow-up questions
Does paying for an AI tool make client-data uploads acceptable?
No automatic permission follows from payment. Review the exact use, data category, tool terms, settings, and current brokerage requirements. A paid account can still be unsuitable for a particular client-information workflow.
Can AI write a listing description?
Use it only through an appropriate process with permitted inputs, accurate facts, required review, and applicable disclosures. Check every material statement against reliable records and remove invented or misleading content before publication.
Can I let AI text clients without reviewing the workflow?
Do not launch unattended communication without understanding consent, identity, disclosure, data use, escalation, and applicable rules. Test with controlled records and define when a human must take over.
Where should I get the required AI disclosure wording?
Use the current company policy and relevant state instructions for the intended communication. Preserve the exact approved wording and placement requirements. This article does not replace the operative policy or authorize a particular tool.
Ask Chris about brokerage fit
Ask Chris a brokerage-fit question about how to evaluate technology workflows and the support needed to use them responsibly.
Sources
- eXp U.S. Policies and Procedures, August 2026 · accessed October 27, 2026
- FTC CAN-SPAM business guidance · accessed October 27, 2026
Fees, programs, and policies change. Confirm current details with eXp's official resources.