ChatGPT Ads for Realtors: Access, Cost and Fair-Housing Rules

Yes, a realtor can advertise on ChatGPT, but as a service, not as a listing board. OpenAI's policy prohibits ads for individual housing sales or rentals. The minimum is $25 a day, and a Clicks campaign bills per click, so a 30-day test starts near $750. Fair-housing advertising rules apply to every word, photograph and targeting choice.

The housing rule is the surprise for most agents, because the obvious first ad, a new listing with a photo, is the one the policy rules out. What remains is still useful: valuation requests, buyer guides, neighbourhood information and your own profile. Everything here is as of early November 2026, in a beta that has changed its policy page several times this year, so read the live policy before you build anything.

Follow me on Instagram@sagnikteaches

What you can put in an ad, and what will be refused

OpenAI's policy sorts advertisers into allowed, restricted and disallowed categories. Real-estate services are not named as restricted, and the policy describes the early focus as consumer categories including local services. The housing rule itself sits under "housing and jobs": ads for individual job listings, or housing rentals or sales, are prohibited. Advertisers may link to a platform that has listings, as long as neither the landing page nor the ad creative refers to a specific listing. Here is how that plays out in practice.

Connect on LinkedInSagnik Bhattacharya

Two more policy points matter. Ads must be consistent end to end, so a service ad cannot land on a page that leads with one specific listing. And the advertiser section says anyone promoting a service that needs a professional licence must keep that licence. Have your licence number and brokerage details ready when the account asks about verification.

Subscribe on YouTube@sagnikteaches

Who sees the ad

OpenAI's help page says ads may appear for users on the Free and Go plans, not Plus, Pro, Business, Enterprise or Edu, and not for accounts identified as under 18. A first-time buyer or a renter who uses the free tier is a plausible viewer. A home owner who pays for ChatGPT for work will not see anything. Before spending, ask your last five clients which ChatGPT plan, if any, they use. If none use a free plan, test a smaller budget or wait.

The cost, worked through for a solo agent

The minimum daily budget is $25 in dollars. OpenAI's help pages suggest a maximum click bid starting at $3 to $5, which is a ceiling rather than the price paid, because ads are chosen in a relevance-weighted, second-price auction. Imagine a solo agent who wants seller valuation requests. Every number except the $25 is assumed.

• Spend: $25 a day for 30 days is $750.
• Clicks: at an average $4 a click, about 187 clicks. At $3, 250. At $5, 150.
• Requests: if 4 in 100 visitors submit the valuation form, 187 clicks give 7.5 requests, so say 7.
• Cost per request: $750 divided by 7 is about $107.
• Appointments: if one request in three leads to a booked appointment, that is just over 2 appointments, so about $320 each.
• Listings: if one appointment in four becomes a signed listing, the month produces roughly half a listing.

Half a listing means one month rarely pays for itself, so the sums have to be run against your own average commission after splits. If a signed listing is worth $6,000 to you, half a listing is $3,000 against $750 of spend. If it is worth $1,500, the same month loses money. This is why a first test should last long enough to produce more than a handful of enquiries, and why the stop rule below uses spend, not feelings. The mechanics of Ads Manager, objectives and billing are covered in the small-business guide to buying ChatGPT ads, so they are not repeated here.

Fair-housing rules in ad copy

The national rule on discriminatory advertisements, 24 CFR 100.75, makes it unlawful to make, print or publish any notice, statement or advertisement about the sale or rental of a dwelling that indicates a preference, limitation or discrimination because of race, colour, religion, sex, handicap, familial status or national origin. It covers written and oral statements and it lists what counts: words, phrases, photographs, illustrations and symbols that convey a dwelling is available or not available to a particular group, expressing a preference to agents or brokers, selecting media or locations that deny segments of the market information about housing opportunities, and refusing to publish advertising or charging different terms. Your state or local law may add protected classes, and your brokerage will have its own policy. I am not a lawyer, so read the regulation and ask your compliance contact.

The practical lesson is to describe the property and the service, never the kind of person who should respond. Some illustrative rewrites:

These are examples of the pattern, not a complete list or a legal safe harbour. The same logic applies to any photograph you attach: people in images, and the choice of who appears, can send a message about who the home is for, and the regulation names photographs and illustrations. If you use ChatGPT to draft variants, give it the rule in the prompt and then read the output yourself. The sibling tutorial on writing listings that pass fair-housing checks gives a fuller prompt for this exact job.

A before and after for a valuation ad

Before
Title: Sell fast with the top local agent
Copy:  Lovely family neighbourhood! Guaranteed best price. Perfect for buyers who want a safe community.

After
Title: Thinking of selling your home?
Copy:  A free written price range from recent local sales, plus a plain guide to the sale steps.
Landing page: [first name], [brokerage], licence no. [number], then the valuation form.

The "before" version has an unsupported superlative and a guarantee, which fail OpenAI's truthfulness rule, plus two phrases that describe the kind of people in a community. The "after" version describes the service and promises something checkable. The identifying details moved to the landing page for a practical reason: OpenAI's launch guide caps an ad title at 50 characters and the copy at 100, and recommends 16 to 24 and 32 to 48. Check your state's licensing rules for what must appear in the ad itself, because a rule that demands the brokerage name in every ad may not leave room for much else. Have the designated broker approve the copy if your brokerage requires it.

Targeting without steering

ChatGPT ads are not keyword-based. At ad group level you write context hints in plain words, and OpenAI says they inform relevance without guaranteeing delivery. You also choose locations and platforms, and standard campaigns can target states, cities and postal codes as well as whole countries. Two fair-housing consequences follow.

First, write hints about the service, never the audience. A hint such as "people thinking of selling a home who want an honest price range" is fine. A hint naming a demographic group is exactly what the regulation treats as a preference. Second, think about location choices. The rule on selecting media or locations that deny some part of the housing market information about opportunities means you should not exclude areas for reasons tied to who lives there. Choose areas by where you work and hold a licence, and document that reason in a one-line note kept with the campaign.

Custom audiences are possible but need at least 25,000 matched users, so an agent's past-client list will not work, and it is better kept private anyway. Do not upload client contact details to any ad platform without reading your brokerage's privacy policy and the platform's terms.

Why creative enhancements need switching off

From 30 October 2026, creative enhancements are on by default for new and existing self-serve campaigns. OpenAI's help page says they can write new headlines and descriptions to suit the conversation, translate the text, resize images and change the destination URL, and that enhanced versions can run without your approval and without a preview of each one. In most categories that is a mild risk. Here, one inserted phrase such as "great for families" could turn a clean ad into a problem, and the publisher is responsible for the final words. Opt out under Manage settings in Ads Manager, at account level and on each campaign, since a campaign's setting overrides the account default. Then view the live ad, screenshot it, and compare it word for word with your approved text. Do that again after any edit and weekly while the campaign runs.

A four-week test plan with a stop rule

1. Before week 1. Confirm licence details, get broker sign-off on two ad variants, and build a valuation landing page that shows no specific listing. Open the account in the right legal name and watch for the sign-up ad that can go live by default; pause it.
2. Week 1. Submit for review. Expect verification and approval to take days, not minutes, and do not fund heavily until the ad shows as approved.
3. Weeks 2 and 3. One campaign, Clicks objective, $25 a day, one ad group, one location set. Do not edit for five days.
4. Week 4. Count form submissions, calls and appointments booked. Note the time you spent replying.

Write the stop rule before spending: "If $750 produces fewer than four valuation requests, I stop and do not retest for a quarter." A rule of that kind protects the budget from hope.

A second ad: a buyer guide, built end to end

Valuation requests suit sellers. For buyers, a guide is the safer first offer because it promises information, not a property. Take an agent who works with first-time buyers. The ad group, the ad and the landing page are drawn up like this. The wording is illustrative.

Campaign: First-time buyer guide (Clicks, $25 a day)
Ad group hints:
- Step-by-step guide for someone buying their first home: deposit, inspection, offer, closing.
- Checklist of the costs between an accepted offer and getting the keys.
Ad title: First-home buying guide
Ad copy: The steps, costs and timings of buying your first home, in plain English.
Landing page: guide download, short agent profile with [first name], [brokerage] and licence no. [number], a form asking for name, email and the month you hope to buy.

The form asks nothing about household make-up, income source or current home. Asking a buyer about their family or background gives you information you do not need and creates a record that could look like screening. Ask only what is needed to send the guide and book a call. If a conversation later needs a budget, take it by phone or in a secure form your brokerage already uses.

The review checklist, filled in for one ad

Each ad should leave a one-page paper trail. Here is an illustrative record for the valuation ad above. It takes about ten minutes and is worth more than any later explanation.

• Ad text: pasted exactly as submitted, with the date.
• Category check: promotes a service, no specific listing, landing page shows no listing. Passed.
• Fair-housing read: no words about who should respond, no people in imagery, no description of the neighbourhood's residents. Passed.
• Licence line: number and brokerage name appear as the brokerage's rules require. Passed.
• Creative enhancements: off. Screenshot of the setting attached.
• Locations: the areas where the agent works, chosen for that reason. Note attached.
• Approved by: agent and designated broker, with dates.

Reporting: what to log and what to ignore

OpenAI says advertisers receive only aggregated information such as views and clicks, so any detail about who responded comes from your own form, not from Ads Manager. Keep a simple sheet with one row per enquiry and these columns: date, source (ChatGPT ad), page the person came from, form type, whether you replied within an hour, whether an appointment was booked, whether it became a listing or a purchase, and the fee or commission if it did. After four weeks, divide spend by each stage. If clicks are plentiful but forms are rare, the landing page is the problem. If forms are plentiful but appointments are rare, your reply speed or the offer is. Views on their own tell you little.

Conversion tracking through OpenAI's pixel or Conversions API is available, and it can count a form submission as an event. Keep it to that single event. Check one pixel feature before launch: OpenAI's documentation describes automatic advanced matching, which, when enabled, detects contact details in recognisable forms on the page, hashes them in the browser and sends the hashes with conversion events. Passing anything someone typed into a valuation form to an ad platform, hashed or not, is a data-sharing decision your brokerage's privacy policy should cover first.

Team names, brokerage rules and who signs off

Many states and brokerages set their own rules about what real-estate advertising must show, such as the brokerage name, the licensee's name, or how team names may be used. Those rules sit alongside OpenAI's policy, not inside it, and an ad that satisfies the platform can still breach them. Before you submit, ask your broker three questions: does advertising on a new platform need written approval, which disclosures must appear in an ad of this length, and does the brokerage keep copies of advertising? Write the answers into the review record. For a team, decide who can edit the Ads Manager account. OpenAI's access guide gives each person an ad account role of Admin, Member or Viewer, and the sensible split is one person who can change copy and a broker or compliance lead who can only view and approve.

A quick sum for the week-two check

At the end of week two, you will have spent roughly $350 and should be able to see a pattern. Say the account shows 90 clicks and 3 form submissions. That is a 3.3 percent form rate and about $117 per request, a little worse than the plan. Do not rewrite anything yet: three submissions is too few to judge. Wait for the full $750, unless a clear fault shows up sooner, such as clicks arriving from a location you did not intend, or an ad that has been rewritten without your approval. Either of those is a reason to pause at once.

If, by contrast, the account shows 90 clicks and no submissions, check the landing page on a phone before blaming the channel. A form that fails to submit, or a page that takes ten seconds to load, wastes the same money as a bad ad. Submit your own test enquiry each week, from a mobile, and confirm that it reaches you and triggers the right reply.

Two things that go wrong in practice

The listing page that sneaks back in. An agent builds a clean service ad, then points it at a landing page that opens with the newest listing. OpenAI reviews ads end to end, so the page can make an approved ad fail. Keep the landing page to the valuation form, the agent's profile and general market information.

Replies that arrive faster than the broker can review. A valuation form that sends an automatic reply written by an AI tool can contain statements about the neighbourhood. Keep automatic replies to confirmation and next steps, and apply the same fair-housing read to every template. The agent's wider routine, including brokerage rules for emails, disclosures and follow-ups, applies here too.

When to wait

If your market is thin and most clients are referrals, the numbers above suggest keeping the money for a channel you already trust. If the main thing you want to promote is a listing, this platform will not do that job at present. If you cannot reply to an enquiry within the hour, do not pay for enquiries. And if your brokerage has an advertising policy that forbids new channels without approval, the approval comes before the account.

Advertise if three things are true: your ad promotes a service, not a property; every word and image has passed a fair-housing read by a human; and one lead is worth several times the $107 to $320 per lead or appointment in the sums above. If any is false, fix that first.

Questions agents ask before a ChatGPT ads test

Can I advertise one of my listings on ChatGPT?

Not as a stand-alone ad. OpenAI's ad policies prohibit ads for individual housing rentals or sales. You may advertise yourself, your service or a general search page, provided neither the ad nor the landing page refers to a specific listing. Treat a "just listed" or "just sold" ad as a risk to avoid until OpenAI's reviewers have said otherwise in writing.

Does the minimum budget apply to a solo agent?

Yes. OpenAI's help pages set a standard minimum daily budget of $25 for dollar-billed accounts, and there is no smaller tier for one-person businesses. A 30-day test is about $750. A daily budget is a weekly average, so on one day the account can bill up to double the figure while the week stays within seven times it.

Do fair-housing rules apply to a paid ad as well as a listing?

Yes. The fair-housing rule on discriminatory advertisements covers any notice, statement or advertisement about the sale or rental of a dwelling, in print, spoken or online, and it names words, photographs and symbols as well as the choice of media or locations. It does not matter that a platform or a tool wrote the first draft. The agent or brokerage that publishes the ad is responsible.

Should I let Ads Manager improve my copy automatically?

No. Opt out of creative enhancements under Manage settings, for the account and each campaign. OpenAI made them default-on for self-serve campaigns from 30 October 2026, and enhanced versions can run without a preview, adding descriptive phrases you have not reviewed. In a category where a single phrase can imply a preference about who may live somewhere, you want to control every word.

Further reads

• ChatGPT Ads for Law Firms: Can You Advertise, and Is It Allowed? — How the same platform treats another licensed profession.
• How to Advertise on ChatGPT: Approval, Budget and Reporting for Agencies — What changes if an agency or freelancer runs the account.
• ChatGPT Bio for a Real Estate Agent: Team Standards, Prompts and Edits — The profile page an ad click lands on needs a good bio.
• Best ChatGPT Plan for Realtors: Free, Plus or Business? — Which ChatGPT plan suits an agent's day-to-day writing.
• Can Estate Agents Be Found in AI Search? What to Fix Now — The free route: being found in AI answers without paying.
• How Estate Agents Use AI to Prepare Valuation Appointments — Prepare the valuation conversation an ad enquiry leads to.
• AI KPIs for Small Businesses: 12 Metrics Worth Tracking — Twelve measures to judge a paid test against.
• AI Real Estate Social Media: A Month of Posts From Your Listings — Plan four weeks of listing and just-sold posts with AI from your own source sheet, with fair-housing and brokerage disclosure checks built in.
• AI Policy for a Real Estate Brokerage: What Agents May and May Not Do — Turn a broker AI policy into rules an agent can follow on a Tuesday: do and don't lists, sample clauses and a sign-off sheet.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: OpenAI's ad policies (housing and jobs section, version history to v1.6, September 2026); OpenAI help centre articles Ads in ChatGPT, Ads in ChatGPT: The Basics, Creating campaigns, Launch Campaigns, Creative enhancements, Conversion Measurement and Managing identity and access for Ads Manager; OpenAI's measurement pixel documentation; the fair-housing advertising regulation 24 CFR 100.75 in the Electronic Code of Federal Regulations; checked 4 November 2026.