ChatGPT Ads for Real Estate Agents: $1,300 Spent, 13 Leads

I spent about $1,300 on ChatGPT ads over the past few months, targeting seller leads here in Austin.

It got me 13 seller leads. That is roughly $100 a lead.

Not $100 per appointment. Not $100 per listing. $100 for a lead, which is an expensive number by any measure I have. I want to put that up front because most of what gets written about new ad channels skips straight to the upside, and the cost is the first thing you would ask me about if we were talking in person.

I am William Zhang, a practicing agent in Austin and the founder of Real Estate AI Society. Below is everything the ad manager actually showed, the three things I genuinely liked, the three things I did not, and whether I think you should be spending money here right now. The short version: the numbers are bad and I am going to keep running them anyway, and the reason for that is the whole point of the post.

The Actual Numbers

Here is what the campaign reported, straight off the screen.

• $1,300 in spend
• 170 clicks
• 13 conversions, which is where the 13 seller leads came from
• 0.83% click-through rate
• About $63 per thousand impressions
• About $7.63 per click

None of that is impressive. A 0.83% click-through rate is low, a $63 CPM is expensive, and $7.63 a click is steep for a channel where I then have to convert that click into a lead and that lead into an appointment.

Run the math forward and it gets worse before it gets better. At roughly $7 to $8 a click you generally need somewhere around 100 clicks to produce a few leads, so you are looking at something like $700 to get two or three. I came out ahead of that pace, which is part of why I think the intent is real, but I would not plan a budget around beating it.

And I have not optimized any of this yet. The ads I ran were a first attempt.

What $100 a Lead Really Means

For comparison, the same seller leads on Facebook run me about $25 to $50, and an appointment there costs something like $300 to $500.

So on cost alone, this channel loses. It is not close. If you put a gun to my head and asked where the next $1,300 goes for the best return this quarter, I would say Google Ads, because it is a more mature system and the lead quality is better dialed in today. I would rather tell you that plainly than sell you a channel I am still figuring out.

There is one more piece of honesty that belongs here, and it is about timing. Most of the 13 are looking to sell three to six months out. A few are closer to right now, but this is not a stack of people trying to list next week. If you need closings this quarter, that matters more than the cost per lead does.

Three Things I Liked

The intent is higher than anything else I have run

This is the part that keeps me in.

A social ad interrupts someone. They were looking at photos of their nephew and now there is a house in the feed. A ChatGPT ad lands inside a question they were already asking, about what their home is worth or what the market is doing or how selling actually works. They are mid-thought about the exact thing I do.

I can also shape when it shows up. I write the prompt targeting so my ads appear when someone is asking about selling a home in the areas I cover, and that is a fundamentally different match than an interest-based audience on Facebook. There is a trust component too. Something the user already treats as a neutral advisor is the surface the ad appears on, and that changes how the message reads.

Almost no agents are in here yet

I could not find much of anything on this. Not many agents running it, and barely any content explaining how to run it for real estate, which is part of why I made the video.

That cuts both ways, and I want to be fair about it. Sometimes nobody is in a channel because the channel does not work. But the pattern with every marketing channel that has ever worked for agents is the same: it is cheap and awkward first, then it is obvious and expensive. The cost here will probably come down, or at least become more reasonable, and the people who already know how to run these ads will be the ones positioned to scale when it does.

The skill is the asset right now, not the leads.

It is going to be a real channel

People are moving their searches out of Google and into these assistants. Not just how-to questions, but purchase research, and increasingly “who is the best person to help me with this.”

I have already seen the organic side of that work. I picked up a client recently who found me through Google’s AI, with no ad spend involved at all, and he was under contract in four days. I wrote up exactly how that happened, because it is the clearest proof I have that this surface already sends real business.

I also do not think ChatGPT will be the only one selling ads. Google has every reason to put advertising inside Gemini once they are satisfied the system is ready, and my honest guess is that lands sooner than most people expect. Claude is harder to call because it is built around paying subscribers, but that tends to shift over time. If you learn how to advertise inside an LLM now, that skill moves to whichever platform matters next.

Right now the ads serve to free users and the roughly $7 a month tier. I picked that up at a ChatGPT ads workshop, and the expectation is that they eventually reach the $20 a month plan as well. With something like 700 to 800 million monthly active users on the platform, even a partial slice is a large reachable audience.

Three Things I Did Not Like

It is expensive just to find out

The usual reward for showing up early to an ad channel is cheap impressions. That is not what is happening here.

The practical floor to get real click volume is about $20 to $25 a day. I know that is not a large amount of money in absolute terms, and I still wanted to test it. But you cannot do what you can do on Facebook, where $5 a day on a good ad can tell you something meaningful. Here you are paying a real minimum before you learn anything at all.

There is not much to work with

The ad itself is a small square image with one line of text, and that line sometimes gets cut off.

No video. No horizontal format. No larger creative. For a channel where the whole job of the creative is to earn a click at $7.63, having one small tile and a truncated sentence to work with is limiting. Standing out is hard when everyone has the same box.

You cannot clearly see what is working

Conversion tracking is not where it needs to be. It is nowhere near what you get from Facebook or Google, and when you cannot trust attribution you cannot optimize with any confidence. You are making changes and hoping.

OpenAI did recently ship a conversion API, which suggests better tracking is coming and the algorithm should improve alongside it. That is the single change I am watching most closely, because it is the one that would let me put real money behind this.

So Is It Worth It?

Both things are true at the same time, and I think agents get this wrong by picking one.

The economics right now are bad. $100 a lead against $25 to $50 on Facebook, on leads that are mostly three to six months out, with creative you cannot really control and tracking you cannot really trust. If you are judging this as a lead source this quarter, the answer is no, and I would point you at Google Ads instead.

But that is not the only question. This is a skill set, and the attention is moving here whether or not the ad platform is ready. Very few agents, teams or brokerages have built any competence in advertising inside an LLM, and the ones who do will have an edge for the next few years while everyone else is still learning the interface.

I am treating the $1,300 as tuition rather than as a lead cost. I will keep testing, keep tightening the targeting and the funnel behind the ad, and share the numbers as they change. If the cost per lead comes down even to the $50 range with this level of intent, the math changes completely.

The other thing worth saying: the ad is the cheapest part of this to fix. Most agents who fail at a paid channel do not fail at the ad, they fail at the page it points to and the follow-up behind it. If nothing is catching those 13 leads and working them for the three to six months it takes, the cost per lead is academic. That is the same argument I make about AI lead generation generally, and it applies here more than usual, because these leads arrive early.

If you want the tools I actually run in my own business, they are all listed here. And I send one AI workflow for agents every week, which you can get here.