AI for Real Estate Agents: Workflows, Prompts and Compliance

AI saves real estate agents the most time on writing: listing descriptions, follow-up emails and texts, social posts and the narrative around a CMA. A general assistant such as ChatGPT, Claude or Gemini handles all four with the right prompts, as long as you feed it facts from the MLS and your own notes and check every claim. The part you can't skip is compliance: fair housing language in every ad, consent before any AI voice call, and disclosure of AI-altered photos where your state requires it.

How agents are using AI in 2026

NAR's 2026 REALTORS Technology Report found that 23% of agents use AI daily and another 25% weekly, while only 12% say they don't use it and have no plans to (NAR, September 22, 2026). Among agents who use AI, these are the most common jobs:

The workflows below follow that list. Each has a prompt to copy and the checks that keep it safe.

Workflow 1: Listing descriptions that stay fair-housing safe

The Fair Housing Act makes it unlawful to publish any ad for the sale or rental of a dwelling that indicates a preference, limitation or discrimination based on race, color, religion, sex, handicap, familial status or national origin (42 U.S.C. 3604(c)). AI drafts often reach for phrases like "perfect for a young family", and the model doesn't know your local rules, so put the rule in the prompt.

Write an MLS listing description of [WORD LIMIT] words for this property, using only these facts: [PASTE MLS DATA AND YOUR NOTES].
Rules:
- Describe the property and its features, never the kind of buyer or neighbor. No phrases like "perfect for young families", "ideal for retirees" or "great for singles".
- Don't describe the people in the area, its demographics or how "safe" it is. Use facts such as distances to parks, transit and shops instead.
- Don't add any feature, upgrade or measurement that is not in the facts above.
- Plain, warm tone. No exclamation marks, no "stunning" or "must-see".
After the description, list every claim you made next to the fact it came from, and flag anything I should double-check.

Then read it against the MLS sheet line by line. Square footage, bedroom counts, school names and HOA details are where AI drafts go wrong most often.

Workflow 2: Lead follow-up in minutes, not hours

New leads go cold fast, and the first reply is mostly typing. Let AI draft it from the lead's own words and your listing facts, then send it yourself.

For leads who don't book, a short nurture sequence keeps you in touch without sounding automated:

Write a five-message follow-up sequence for a buyer lead who asked about [PROPERTY OR AREA] and has not booked a showing. Their words: "[PASTE THEIR MESSAGE]".
Mix email and text across 3 weeks. Each message gives something useful: a similar listing, a price update, an open house or a question about their timeline. Under 70 words each, in my tone: [DESCRIBE YOUR TONE].
Every email ends with my brokerage name, a physical address line and an unsubscribe sentence. Never invent listings or prices: use only [PASTE CURRENT LISTINGS].

Three rules apply whatever tool sends the messages:

• Email: CAN-SPAM requires honest subject lines, a valid physical postal address, a clear way to opt out, and opt-outs honored within 10 business days (FTC compliance guide).
• AI voices: the FCC ruled in February 2024 that AI-generated voices count as "artificial" voices under the TCPA, so those calls need the called person's prior express consent (FCC ruling).
• Cold outreach: Do Not Call rules still apply to cold calls, whether a person or a tool dials.

Missed calls are their own workflow. Our guide to building an AI receptionist covers what it should and shouldn't say, and when to hand off to you.

Workflow 3: Social posts and listing media

Social posting is the second most common AI use among agents. Batch it: one session a week produces the whole week's posts.

Plan 5 posts for next week for a real estate agent in [MARKET]. Use these inputs: new listings [PASTE], recent sales [PASTE], one local event [EVENT], and one market fact from [SOURCE AND FIGURE].
For each post: the platform, the hook in under 12 words, the caption under 90 words, a photo or video idea, and a call to action.
Rules: describe homes and places, never the ideal buyer. No guaranteed returns or price predictions. Include my brokerage name where my state requires it: [YES OR NO].

To automate the scheduling as well, see how to automate social media posts with AI.

Photos need their own care. California's AB 723, signed in October 2025 and in force since January 1, 2026, requires agents who use a digitally altered image in a sale listing, including AI virtual staging or removing poles and furniture, to add a disclosure on or next to the image and a link or QR code to the original, unaltered photo (bill text). Lighting, color correction and cropping don't count. Outside California, NAR's Code of Ethics still requires a "true picture" in advertising, so disclose staging even where no law forces you to.

Workflow 4: CMA prep without handing over the pricing

AI should never choose comps or set a list price. You pick the comps in the MLS; AI does the arithmetic and the writing.

Attached is a spreadsheet of comparable sales I selected for [SUBJECT PROPERTY ADDRESS, BEDS, BATHS, SQUARE FEET].
1. For each comp, calculate price per square foot and days on market. Show the formula.
2. Give the median and the range, and flag any comp more than 15% from the median with a possible reason from the data.
3. Draft a one-page explanation for the seller of what the comps show, in plain language.
Do not suggest a list price. I will set it after reviewing your math.

Before you upload an MLS export to any AI tool, check your MLS's rules on how its data may be used, and remove owner names and contact details you don't need. For a spreadsheet with many comps, or a folder of past CMAs, an agent such as Claude Code works directly on the files; Claude Code for non-developers shows how.

The tools that cover most of it

Claude Pro is $20 a month billed monthly (Anthropic pricing), and most assistants have a similar paid tier. Start with one assistant and add tools only when a workflow proves itself; AI for small business covers how to roll AI out across a whole office.

Your compliance checklist before anything goes live

What to do next

Pick the one workflow that eats the most of your week, probably listing descriptions or follow-ups, and run its prompt on your next real listing or lead. Check the output against the MLS and the compliance checklist, then save the prompt with your edits so the next one takes minutes. If you want to go further, the ai school courses teach Claude Code for batch work, landing pages and AI video, with an automation course for follow-up in early access.