AI for Real Estate: Where to Start

AI for real estate works best on three jobs: first drafts, fast answers to routine questions, and paperwork summaries. Pick one task, test one tool for four weeks, check every output before clients see it, and keep AI out of decisions about who gets housing.
If you want to talk through your own case, see the AI consulting page. Below: what AI can do, who is using it, a four-week start, and the rules that matter in this industry.
What can AI do for real estate?
For most agents and property managers, AI is a fast first-draft writer and a fast first responder. It does not close deals or inspect buildings. The table lists the common jobs and how risky each one is. The ratings are my judgment, not a study, so test before you trust them.
| Job | Who it helps | Effort to start | Risk if it is wrong |
|---|---|---|---|
| Draft listing descriptions | Agents, property managers | Low | Medium: facts must match the property, and the wording is covered by advertising rules |
| Draft emails, texts and follow-ups | Agents, property managers | Low | Low if you read it before sending |
| Answer common questions on your website or by chat | Agents, property managers | Medium | Medium to high, because it speaks for you |
| Summarize contracts, leases, inspection reports and call notes | Everyone | Low to medium | Medium: a summary can leave out the clause that matters |
| Edit listing photos or stage rooms virtually | Agents | Low | Medium: disclosure rules and the risk of misleading buyers |
| Sort maintenance requests and route them to a vendor | Property managers | Medium | Medium |
| Research comparables and market data | Agents, investors | Medium | High if you rely on numbers you did not check |
| Screen applicants or decide who sees an ad | Property managers, agents | Not a first project | Very high: fair housing law applies |
The first two rows are where nearly everyone should begin. They are easy to check and easy to measure.
How many real estate professionals use AI?
Quite a few, and most say it is helping a little. In the 2025 NAR Technology Survey, released September 18, 2025, 20% of REALTORS said they use AI daily, 22% weekly and 27% a few times a month, while 32% had not yet used it in their business. Asked about the effect, 17% reported a significantly positive impact, 33% a moderately positive impact and 46% no noticeable impact. ChatGPT was the most used tool at 58%, ahead of Gemini at 20% and Copilot at 15%.
Two cautions. This is a survey of REALTORS, so it says little about property managers or investors. And the answers are self-reported, so “positive impact” means people felt it helped, not that anyone measured it. The useful reading: most of the industry is still early, and almost half of the people who tried it saw no difference. The ones who did see a difference usually picked a narrow job.
Where should you start with AI in real estate?
Follow five steps, in order. Skip the first and you have nothing to measure the other four against.
1. Pick one task that repeats
Choose something you do every week that is mostly writing or routine decisions, where a wrong answer is cheap to catch. Good first picks: first drafts of listing descriptions, replies to the questions every lead asks, follow-ups after a showing, summaries of call notes. Write the task and the number it moves in one sentence: “Cut the time I spend writing follow-ups from six hours a week to three.”
2. Measure speed to reply, not only hours saved
Speed matters here more than in most businesses. It is a game of who gets back to the lead first. Reply within minutes and people think that’s impressive, and that is probably the level of service they expect from you. The same goes for chat on your website. Visitors ask what something costs, and if a person gets back to them four minutes later, they have already bounced. That is why I thought a chatbot that could pull answers from the site would be a game-changer for support and sales. So track how long a new lead waits for a first reply, before AI and after.
Follow-ups are an easy place to start. I let Claude draft my own email replies in my voice, and it only drafts: I read each one and press send myself, which is the same read-every-draft rule I use below. I show the setup in my email assistant video.
Video · 14 min
How I Let Claude Automate My Email Replies (For Gmail & Microsoft)
I show how I have Claude draft email replies in my voice, set up for Gmail or Google Workspace, and it never sends anything for me.
Watch the video →
3. Pick one tool
Pick one, not three. Check the software you already pay for first, because many CRMs, listing tools and property management systems now include an assistant. If none covers your task, choose one general assistant and use the cheapest plan that lets you test it. Before you paste anything real, read what the vendor says about your data on that exact plan. If you would rather have a website chat or follow-up tool built for you, I build custom AI apps that you own.
4. Run it for four weeks and read everything
Give one person ownership. Write down the instructions you give the tool so a colleague can repeat them. Read every draft before it reaches a client, a tenant or the public. The four-week length is a practical suggestion, not a research finding: it is long enough to see the work repeat and short enough to stop.
5. Decide with the numbers
Compare hours, reply times and how often a draft needed a rewrite. Set the stop rule in advance: “If more than one draft in four needs a full rewrite, we stop.” I treat ideas as assumptions to test, not conclusions, and stopping is a result, not a failure. If it worked, give the same tool a second, similar task. Do not add three new tools at once.
For prompt habits that make drafts better, see AI prompts.
What should AI never do in real estate?
Keep AI away from any decision about who can rent, buy or see a home, and keep a person in the room where trust is built. Two separate reasons.
Fair housing. On May 2, 2024, HUD released two guidance documents saying the Fair Housing Act applies to tenant screening and to the advertising of housing, including when algorithms and AI do the work. HUD’s release also warns that ad targeting tools can break the law, for example by withholding housing information based on protected traits. An owner or manager can be responsible even when a third-party tool does the screening or targeting. The practical rules:
- Do not use AI to screen, score or rank people in ways tied to a protected class.
- Do not use AI to choose who sees a housing ad, or to describe a neighborhood by who lives there.
- Have a person review listing copy for wording that hints at who should or should not apply.
- Check your local and state rules too, which can go further than federal law. HUD’s guidance is from 2024, so confirm its current status before you rely on it.
The human part. Deals aren’t completed through written communication alone. Sooner or later you have to talk face to face with clients, and ChatGPT won’t be there to coach you on what to say. Use AI to prepare, draft and summarize, then do the conversation yourself.
Which rules keep changing, and how do you stay current?
Advertising, photo and contact rules move often, so build a habit of checking them. I have written hundreds of blog posts on topics where the rules keep changing. I bring in attorneys to review them and join webinars, and you have to check everything all the time and stay on top of it. The same habit applies here.
Three places to check first:
- Listing photos. If you use AI to stage a room or enhance a photo, your MLS may have a rule. As one example, NorthstarMLS posted proposed guidelines on July 10, 2026 asking that any virtually staged, AI-generated or AI-enhanced photo be labeled and shown with an unaltered “before” image. Your MLS may differ, so read its own rules.
- Calls and texts. Before an AI sends outreach on your behalf, ask your broker or a lawyer about consent rules for calls and texts.
- Fair housing and advertising. Your state real estate commission and your broker’s compliance office are the place to ask, not a chatbot.
What changes by role: agent, property manager, investor?
The method is the same. The first task differs.
- Agents: start with listing drafts, lead replies and showing follow-ups. The bottleneck is speed and consistency.
- Property managers and landlords: start with replies to common tenant questions, maintenance request summaries and draft notices that a person approves. I stay at the tools-and-workflow level here: I co-founded DoorLoop and was its CMO until I left in July 2026, but I am not a property manager myself. Keep AI away from screening and anything that decides who gets a unit.
- Investors: start with summarizing documents and organizing your own notes. Treat any number AI gives you for a comp or a return as unchecked until you have traced it to a source.
How do you choose an AI tool for real estate?
Ask six questions of any tool, whether it is a general assistant or a product built for the industry.
- Which one task does it do better than what I do today?
- What happens to the client data I put in, on this exact plan?
- Can a person review and edit every output before it goes out?
- Does it ever screen, score or target people, and if so, how does the vendor handle fair housing?
- Can I leave, and take my data with me?
- Can I try it on one task for a month before I commit?
If a vendor cannot answer the fourth question in plain words, treat that as the answer. For the wider checklist, read AI agents for business and AI governance.
What are the next steps?
- Write the one task and the number it should move.
- Record the baseline: hours per week and minutes to first reply.
- Pick one tool and read its data terms.
- Run four weeks, reading every output.
- Decide with the numbers: widen, change or stop.
- Write a one-page rule for your team on what never goes into an AI tool and what AI never decides.
Related guides
- AI for small business: where to start, the general version of this four-week method
- AI governance, for the team rule on data and approvals
- AI prompts, for better first drafts
- AI agents for business, for chat and follow-up automation
- AI risk, for the risks to handle before you widen
If you want a second opinion on your first task or your plan, start on the consulting page. For team training, see my AI workshops.
Updated November 14, 2026. Sources opened on that date: the NAR 2025 Technology Survey release, the HUD news release 24-098 and the NorthstarMLS guidelines page.


