AI earns its place in a real estate business where time leaks: first response to a lead, follow-up past the second attempt, CRM notes, listing drafts, and website questions after hours. It does not earn a place in advice, negotiation or anything sent without consent. The difference between a chatbot that answers and an agent that acts is explained in AI agents vs chatbots.
This guide is part of our real estate marketing cluster. What a chatbot costs is in how much a chatbot costs; the brokerage-level plan these tools serve is in the real estate brokerage marketing plan.
Where AI helps an agent, and where it does not
The honest split:
- Helps: first response to an inquiry within minutes, follow-up sequences that continue past the second attempt, call and meeting summaries into the CRM, listing descriptions and posts drafted from facts, website chat that answers from listing data and books a showing, and sorting the inbox.
- Does not help, or should not: pricing advice, negotiation, anything that reads as legal or financial advice, sending to anyone who has not consented, and publishing anything a person has not read. These are the agent's judgment and the agent's liability.
Use 1: follow-up
The first job is speed: an inquiry from the website, a portal or a sign acknowledged within minutes, with a real next step, at any hour. The second is persistence: a sequence of useful messages over weeks for leads who are not ready, drafted by AI from the lead's situation and sent only within the consent the lead gave. The third is the CRM: calls and meetings summarized, next actions set, notes written, so that the agent's memory is not the database. All three are automation around the agent, not instead of the agent; our AI automation work builds them into the CRM a brokerage already uses.
The rules sit here. In the USA, 47 CFR 64.1200 prohibits calls using an automatic telephone dialing system or an artificial or prerecorded voice to a cellular number without the prior express consent of the called party, and where such calls include an advertisement or constitute telemarketing they need prior express written consent: a signed agreement that clearly authorizes the seller to deliver such calls to the stated number and discloses that signing is not a condition of purchase. The rule applies its solicitation provisions to text messages to wireless numbers, bars telephone solicitations before 8 a.m. or after 9 p.m. in the called party's time zone, and requires national do-not-call registrations to be honoured indefinitely. In Canada, section 6 of the anti-spam legislation prohibits a commercial electronic message without express or implied consent and requires identification and an unsubscribe mechanism; section 10(10) implies consent for two years after a purchase and six months after an inquiry, and section 11 requires unsubscribes honoured within 10 business days (text current to 19 October 2026). An automated sequence must read the consent record before every send.
Use 2: listings and content
AI drafts well from facts: give it the property data, the neighborhood facts and the brokerage's voice guide, and it returns a description, a set of social posts and an email in minutes. It does not know what is true, so the agent supplies the facts and checks the output; a draft that invents a renovated kitchen is the agent's false statement once published. The REALTOR Code's Article 12 requires honest and truthful communications that present a true picture, and SoP 12-10 prohibits misleading images, which includes AI-edited or generated photographs presented as the property.
The second check is Fair Housing. The Fair Housing Act prohibits housing discrimination on the basis of race or color, religion, sex, national origin, familial status and disability, and SoP 10-3 of the Code bars any advertisement indicating a preference, limitation or discrimination. AI models can produce phrases that describe the buyer rather than the property ("perfect for young families", "walking distance to the church"); the brokerage's prompt should forbid them and a person should read every draft for them.
Use 3: website chat and showings
A chatbot that answers from the listing data (price, size, taxes, schools by distance, showing times) and books a showing or a call with a named agent handles the questions that arrive after hours, which is when buyers browse. The rules for it: it says it is a bot, it answers only from the data it is given and says when it does not know, it hands over to a person on request and for anything that is advice, and it collects consent before it sends anything afterwards. Our chatbot development service builds to that pattern; the difference between a bot that answers and an agent that acts on systems is in AI agents vs chatbots.
Use 4: paid ads, with the housing restrictions
AI can draft and test ad copy and pick images, but the targeting rules are fixed by the platform and the law. Meta's special ad categories fix housing ads to ages 18 to 65+, allow no gender selection and no lookalike audiences, require location selections at or above a 15-mile or 25-kilometre radius in the US and Canada, and restrict interests to an approved list. No AI tool may route around that, and an ad's copy is held to the same Fair Housing and true-picture standards as a listing. Our guide to Facebook ads for real estate covers the campaigns.
What AI costs to run at agent scale
The model is the small part. OpenAI's API pricing (standard tier, read 8 November 2026) runs from $0.10 per million input tokens for its smallest current model to $10 for its largest, with mid-range models at $1 to $2 in and $5 to $10 out; an agent's monthly drafting and chat volume is a few dollars to tens of dollars. The larger costs are the setup (connecting the CRM, the listing feed and the consent records), the chatbot's build or subscription, covered in how much a chatbot costs guide, and the agent's time reviewing every draft, which is not optional.
What to tell clients, and what not to claim
Say that you use AI to draft and to answer first, and that a person reviews what goes out and takes over on request. Do not claim the tool does what you have not tested. The FTC's advertising rules require advertising to be truthful and non-deceptive and advertisers to have evidence to back up their claims, and its Operation AI Comply announcement of 25 September 2024 named a company that advertised itself as the world's first robot lawyer without testing whether its output matched a lawyer's, with the Chair's statement that using AI tools to trick, mislead or defraud people is illegal. A brokerage that advertises "AI that sells your home" should be able to show what that means.
A plan for one brokerage
Six steps, in the order that keeps the risk low:
- Fix the consent record first. Every contact in the CRM with its consent type, date and source; nothing automated sends without reading it.
- Automate the first response and the CRM notes. The two leaks with the quickest return and the least risk.
- Write the voice and the forbidden-words guide, including the Fair Housing phrases AI must never use, and give it to every tool and every agent.
- Add drafting for listings and posts, with the rule that the agent supplies the facts and reads every draft.
- Add website chat answering from the listing data, with hand-over and consent capture.
- Review monthly: response times, follow-up continuation, complaints, unsubscribes and any wrong answer the bot gave.
Common mistakes
The faults we see most:
- Sequences that ignore consent. A text to a mobile number without prior express written consent is the costliest message an agent can send.
- Publishing unread drafts. The invented feature or the Fair Housing phrase is the agent's statement.
- A bot that pretends to be the agent. Say it is a bot; hand over on request.
- Letting AI set the price. Pricing is advice; it is the agent's.
- Claims about the AI you cannot show. The FTC treats AI claims like any other claim: true and backed.
- No monthly review. Wrong answers and complaints compound quietly.
Conclusion
AI for real estate agents is follow-up, listing content and website chat, automated around the agent and never without consent or review. The rules do not change because software sent the message: the TCPA and CASL decide what may be sent and to whom, the Fair Housing Act and the REALTOR Code decide what may be said, and the FTC's rules decide what may be claimed about the tools themselves. Fix the consent record, automate the leaks, keep a person on everything that goes out, and review monthly. This guide explains the rules as published; it is not legal advice.
PromotEdge International is an international strategy, technology, marketing and design company, and our delivery team is in India. Our chatbot and AI automation teams build follow-up, drafting and chat for brokerages this way, and our real estate marketing team runs the channels they serve.
Facts we rely on: see our company facts for what we state about ourselves and where each fact comes from.