Back to Notes
AI & AutomationNovember 16, 202611 min read

ChatGPT Ads for Real Estate: My $655 Test Results

In Poland, I spent $655 testing ChatGPT ads for a housing development. The results: 55,735 impressions, 882 clicks, and 0 leads in September 2026.

A laptop displays an AI chat window with a sponsored ad for townhouses, next to a small white house model and a phone showing an analytics chart, with no readable text.

More than 55,000 ad impressions, 882 clicks to the website, and 0 leads. That was the result of our first month advertising a housing development near Warsaw, Poland, through ChatGPT.

We spent about $655 (exactly 2,519.49 PLN) on the campaign in September 2026. I expected OpenAI's new advertising platform to work particularly well for real estate. After all, people talk to ChatGPT about mortgages, moving, choosing a home, and the cost of owning property. It's hard to imagine a better context for advertising a housing development.

But being interested in a topic doesn't mean someone is ready to buy. I learned that with my own advertising budget.

Why Did I Decide to Test ChatGPT Ads for Real Estate?

In August 2026, OpenAI announced that it was expanding ChatGPT ads into European markets. Starting August 31, 2026, businesses could launch campaigns themselves through Ads Manager.

I run REND.PRO, a real estate marketing company in Poland. So when the opportunity to test ChatGPT ads became available here, I didn't want to judge the platform based on announcements or what other marketers were saying.

We selected a real housing development near Warsaw, Poland, and planned a September 2026 test with a budget of about $780 (3,000 PLN). The goal was simple: drive people interested in the properties to the website and see whether they would start contacting the sales office.

My expectations came from how differently these platforms work.

On Google, someone might search for "houses for sale Piaseczno," referring to a suburb of Warsaw. That's a very specific signal. On Facebook or Instagram, an ad often appears before someone has even started actively searching.

ChatGPT seemed to sit somewhere in between. People describe their plans, needs, and limitations in conversations. In theory, that could help the system understand what they're looking for more accurately.

It was an attractive hypothesis. But only the campaign results could tell us whether that understanding would translate into contact from potential buyers.

Setting Up 10 ChatGPT Ads: The First Problem

In September 2026, OpenAI's advertising dashboard was still in beta. I tried to upload multiple ads at once, but the bulk upload feature didn't work.

In the end, I created 10 ads manually. ChatGPT helped me prepare the headlines, descriptions, names, and campaign parameters. I added the images myself.

We also configured OpenAI's measurement pixel and added UTM tags to our links. This allowed us to separate traffic from ChatGPT from visits coming through other channels.

I didn't want to evaluate the entire experiment based on impression counts inside the advertising dashboard.

The upload issue didn't determine the campaign's performance. But it was a reminder that we were working with a very new platform.

That wasn't enough to make me quit. If a campaign brings in valuable inquiries from potential buyers, I'm perfectly happy to set up 10 ads manually.

The question was whether it would.

How Much Did ChatGPT Ads Cost, and What Were the Results?

After September ended, we reviewed the advertising report.

MetricOpenAI Ads, September 2026
Planned budgetAbout $780 (3,000 PLN)
Actual spendAbout $655 (2,519.49 PLN)
Ad impressions55,735
Clicks to the website882
Average CPCAbout $0.74 (2.86 PLN)
Started contact forms0

It's not that the ads were invisible. They generated impressions and brought people to the website.

The problem came afterward.

According to the report, not a single website visit attributed to OpenAI Ads resulted in someone starting a contact form.

I'm not saying the ads generated zero sales, because the report doesn't establish that metric. I'm also not saying nobody was interested in the development.

What I know is how many measured clicks to the website we received and how many contact forms were started. Those are the numbers I'm using to evaluate the campaign.

For a campaign designed to generate inquiries from potential homebuyers, this was a poor result.

Website traffic alone, even when it's relatively inexpensive, doesn't pay for another month of advertising.

ChatGPT Ads vs Google Ads: Same Development, Same Month

Here's what made the test particularly interesting.

During September, we were also advertising the exact same housing development through Google Ads and Meta Ads.

That meant I could compare results from the same period instead of looking at a Google campaign from several years earlier.

MetricOpenAI AdsGoogle Ads
SpendAbout $655 (2,519.49 PLN)About $1,760 (6,765.79 PLN)
Impressions55,73532,626
Clicks to the website8822,662
Average CPCAbout $0.74 (2.86 PLN)About $0.66 (2.54 PLN)
Started contact forms015

Google delivered fewer impressions, but more clicks to the website and 15 started contact forms.

That distinction matters: 15 started contact forms do not mean 15 submitted inquiries or 15 customers.

Across the entire website, we recorded 12 submitted forms from all traffic sources combined. There's no reason to treat these different metrics as the same thing.

Meta Ads spent about $308 (1,183.60 PLN) during the same period and generated 682 visits to the website.

However, the report doesn't provide a comparable number of valuable contacts attributed to Meta. I'm not going to credit that platform with a result I can't verify.

This also wasn't a controlled A/B test with identical budgets, audiences, and ad formats.

We can't use these numbers to calculate some universal advantage of Google Ads over ChatGPT ads.

What we can do is make a practical decision about this particular housing development and how to spend its advertising budget.

There's one more detail worth mentioning.

The report showed an average time on site of about 15 seconds for OpenAI traffic and about 30 seconds for Google traffic.

Time on site alone doesn't tell us whether someone was interested in buying a property. But it's another signal that visitors from the two channels behaved differently.

ChatGPT Understands Conversations, But Can It Identify Buyers?

This is where we get to the central question behind the entire experiment.

Imagine two people.

The first searches Google for "townhouses for sale Piaseczno," a suburb of Warsaw. They're looking for a specific type of property. They might already be comparing prices and locations.

The second asks ChatGPT, "Is it better to buy a house or an apartment?"

Maybe they'll start looking for property next year. Maybe they just want to compare the costs. Or maybe they have absolutely no intention of buying anything.

Both people are talking about real estate.

But the topic of a conversation alone doesn't tell us how close someone is to making a purchase decision.

I think this is the easiest mistake to make when evaluating advertising based on conversational context.

We can get excited about how well AI understands language and forget what an advertising campaign is supposed to accomplish.

The goal isn't simply to find someone who mentions a house.

It's to reach someone who might actually take the next step toward buying one.

I don't have access to the conversations of people who saw our ads. Advertisers don't receive users' private chats or conversation histories.

OpenAI also explains that ads are separate from ChatGPT's responses. Paying for a campaign doesn't make the assistant recommend the housing development you're advertising.

So I can't prove that our ads were mostly reaching people who had no intention of buying.

That's my hypothesis, not an established explanation for the campaign's poor performance.

But after seeing the report, it's the first thing I would investigate.

Are ChatGPT Ads More Like Google Ads or Meta Ads?

Before running this test, I was tempted to think of ChatGPT as a better version of Google.

In practice, I think it's also worth looking at ChatGPT advertising through the lens of Meta.

Google Search often reaches people who actively express a need through their search query.

Meta is more commonly used to build interest earlier, while people are browsing content.

A ChatGPT ad can appear while someone is looking for answers or considering different options. But not every conversation about a product or topic represents purchase intent.

That's why Google was my benchmark for evaluating business results, while Meta provides a useful comparison for the role this new advertising channel might play.

This doesn't mean every ChatGPT campaign is a brand awareness campaign.

OpenAI is developing its targeting, measurement, and conversion optimization capabilities.

My point is simpler.

A new advertising format doesn't automatically become more effective just because it appears alongside an intelligent conversation.

The quality of an AI assistant's answers and the effectiveness of its advertising platform are two different things.

Why I Stopped Testing ChatGPT Ads After One Month

I have nothing against experimenting.

Actually, I like trying new technologies early, before everyone starts talking about them at marketing conferences.

But I'm not going to keep funding a campaign just so I can say my business is advertising with AI.

After one month, the business situation was clear enough.

OpenAI Ads was generating website traffic, but it wasn't producing the measured first step toward contacting the sales office.

At the same time, Google Ads was generating started contact forms, and its reach was constrained by the daily advertising budget.

So I ended the experiment.

The recommendation in our September report was to turn off OpenAI Ads and move the planned budget to Google.

This isn't a final verdict on ChatGPT advertising.

We ran the test at a very early stage of the platform's availability in Poland.

Results could look different in other industries, with other offers, in future months, or when the advertising tools become more mature.

In my specific test, the results didn't justify spending more money.

And that was enough to make the decision.

What I Would Check Before Spending Money on ChatGPT Ads

If I were evaluating a similar experiment today, I'd start with three questions.

First: Why am I running this campaign?

If my goal is to understand a new advertising channel or reach people who are starting to explore a topic, I would evaluate the campaign differently from one designed to generate direct sales inquiries.

Second: What am I actually measuring?

Impressions, clicks to the website, started contact forms, and submitted forms are different stages.

I wouldn't group them all together under one vague label.

Before launching, I'd make sure the UTM tags were in place and that the relevant events were being tracked correctly.

Third: What will I do with the results?

It's worth deciding in advance when a campaign deserves another chance and when the budget should go back to a channel that's already producing measurable actions.

Without those rules, it's easy to convince yourself that a new dashboard and an interesting advertising format have value simply because they're new.

I'm open to the possibility that ChatGPT ads will become much more effective.

I have no reason to want this technology to fail.

I'd genuinely like to see better results as the platform matures.

But my September 2026 experiment ended with very specific numbers: about $655 (2,519.49 PLN) spent, 882 clicks to the website, and 0 leads.

We got impressions. We got clicks.

What we didn't get was any contact activity attributed to that channel in the metric we were tracking.

If you were spending your own advertising budget, would you give this campaign another month, or move the money to a channel that's already generating inquiries?

Sources and Measurement Notes

The campaign data comes from our internal September 2026 advertising report for the same housing development near Warsaw, Poland.

The report uses different metrics, including impressions, clicks to the website, average cost per click, started contact forms, and submitted forms.

The numbers refer to these specific advertising campaigns, not the overall advertising market.

For official information about the European launch of ChatGPT advertising, see OpenAI: ChatGPT ads expand across Europe.

For details about how ads work in ChatGPT, including the separation between ads and assistant responses and the privacy of user conversations, see OpenAI Help: Ads in ChatGPT.

The platform's features and capabilities may change over time.

PLN converted to USD at the National Bank of Poland rate of September 30, 2026: 1 USD = 3.8449 PLN.

Paid online consultation

You can book a paid online consultation. We will review your situation, identify practical next steps and I will tell you directly what I think is worth doing next.

No fluff. A practical conversation with clear takeaways and a founder's perspective on building online projects.

The consultation is held online. We will confirm the details and time by email.

Discord
alexandergadomski