A Third Kind of Ad Intent
Paid search was built around a query. Social was built around an audience. The ads showing up inside ChatGPT are built around something else: a decision in progress. For real estate and lending, that's a moment the other two channels can't reach.
Paid search was built around a query. Social was built around an audience. The ads now showing up inside ChatGPT are built around something I have not had to plan for before. A decision in progress.
That is a genuinely new place to put an ad, and it sits next to the two channels most real estate and lending marketers already know cold.
Two kinds of intent we already know
Google search is stated intent. Someone types “mortgage rates” and you meet a decision that has already started. The person told you what they want, so you show up and serve it. It is efficient because the query does the hard part.
Meta is an audience you interrupt. Nobody scrolling their feed was looking for a lender. You introduce yourself mid-scroll and try to plant something in someone who was not in the market this morning. I have written about these two as demand capture and demand creation. They sit at opposite ends. One harvests intent that already exists. The other tries to manufacture it in someone who was not looking. Both are useful. Neither describes what happens inside a chatbot.
The third position
ChatGPT ads sit in a spot neither of those covers. The person is not typing a two-word query, and they are not passively scrolling. They are in a back-and-forth, working a problem out loud with an assistant.
A borrower starts with “can I qualify if I am self-employed.” Ten minutes later the conversation has moved through write-offs, debt-to-income, bank statements, down payments, and “should I just talk to someone about this.” That whole thread carries far more about the person’s real situation than the keyword self employed mortgage ever could. The ad is reaching them in the middle of reasoning it out. Not before the thought, the way search does, and not adjacent to no thought at all, the way a feed does. During it.
Why housing fits this better than almost anything
Nobody buys a house or takes out a loan off a single search. They research for months. They compare, they run the numbers, they second-guess, they ask the same question five different ways until it finally makes sense.
That is precisely the kind of long, messy, question-heavy thinking an AI is now hosting. A first-time buyer, a relocation, a self-employed borrower, someone deciding whether to sell before they buy: each one is a conversation, not a click. So this surface can put you in front of high-consideration buyers while the whole accumulated context of the conversation is in play. A single query carries almost none of that context, and a feed interruption carries none at all. The person may be at the same point in the decision either way. What is different is that the platform can match your ad to everything the conversation has surfaced, instead of to two words or to nothing. The decision is playing out in the conversation, and for the first time there is a way to be relevant to it.
Where the ad sits, and why that is a strength
One detail matters more than it looks. In what OpenAI has been testing, the ad shows up below the assistant’s answer, kept separate from the answer itself. You also never see the conversation yourself. The platform does the matching on its side; you bring an offer and a sense of who it is for, and you get placement next to the right kind of moment, not a transcript of it.
That line is worth respecting rather than resenting. The AI is not recommending you. It answered the person’s question honestly, on its own, and then an ad appeared next to a problem it just helped clarify. That is a cleaner deal than it sounds. The organic answer keeps its integrity, which is the only reason people trust the assistant enough to be there in the first place. And your ad gets to meet someone whose problem was, thirty seconds ago, sharpened for them by the very thing they were talking to. That is a good moment to show up in.
The part that makes me cautious
There is a catch I keep coming back to, and it is specific to housing.
Being present in someone’s private reasoning is a kind of trust, and trust is easy to break. If the ad feels like the assistant just sold the person out, it backfires harder than a bad Google ad ever would, because the person was being vulnerable, thinking out loud about money and about where they are going to live. You are a guest in that moment. Guests who overreach do not get invited back.
Housing and lending also carry rules that most advertisers do not think about until they get burned. Fair-housing law is built around protected characteristics, things like race, religion, national origin, sex, disability, and familial status, and it reaches the whole campaign, not just the words in the ad but who you target and who the system ends up delivering it to. A channel this rich in personal context is exactly where it is easy to cross that line without meaning to, so I would go in deliberately. Use the context to be relevant to the decision, not to sort people by who they are. The platform layers its own rules on top of that, and regulated categories like mortgages are gated, so what you can even run is not a given. That distinction is most of the job here, and it is one more reason this is a place for judgment, not just a bigger budget.
What this is not
This is a new position on the board, not a replacement for the other two. The platform is young, it is changing month to month, and it is tiny next to Google and Meta.
I would not move a real budget onto it, and I am not going to pretend the numbers are proven, because for now they are not. But paid search took years to become the obvious thing everyone does, and the people who understood its intent early had a long head start on everyone who piled in once it was obvious. A third kind of intent is worth understanding while it is still early, before it is a line item every competitor is fighting over.