OpenAI says it will start testing ads in ChatGPT in the U.S. later this month, and the first version is small: labeled visual ads that show up while the tool is making an image. Small first versions are how big shifts start. Here is what was announced, what was not, and what I would do about it if I owned a business or simply used these tools every day.
What OpenAI actually announced
On October 5, 2026, OpenAI said it will begin a U.S. test of visual ads with a first group of advertisers later in October. The ads appear during image generation. They sit separate from the image you asked for, and they are clearly labeled as ads. As far as placement goes, that is the whole announcement.
OpenAI also says the ads do not influence ChatGPT's answers. Hold onto that sentence, because the entire conversation turns on it. By the company's own description, you ask a question, you get an answer, and a paid message does not shape that answer. The ads live in their own lane, in one part of the product, in one country, as a test.
The reach is why anyone cares. OpenAI reports 1.2 billion weekly ChatGPT users. I read that number on the Tuesday LIVE right after our Santa Clarita escrow count. Every advertiser on the planet wants to stand where 1.2 billion people walk past every week. If you work in marketing, you have probably already started reading the fine print and lining up a budget.
What the announcement does not say matters just as much. It does not say ads appear inside answers. It does not say a brand can buy its way into a recommendation. It does not say the test is worldwide. If you see someone online claiming any of those things, ask them where they read it, and wait for a real answer before you repeat it.
Why the answers feel neutral right now
I had this exact conversation with a client earlier that Tuesday, before I ever went live. We were talking about how the assistants behave today. You ask for help, and the tool gives you options, a little opinion, a few pros and cons. It rarely plants a flag.
"Usually, it kind of does it in a noncommittal way, right? It's not in love with the things it's suggesting."
That shrug is a big part of why people trust these tools. A salesperson leans toward the thing he wants you to buy. A reference book does not care which page you turn to.
Right now the assistant feels more like the reference book, and that feeling has real value. It is also fragile, because you cannot see what sits underneath an answer. You are trusting the tone.
Here is the example I used on air. Say you tell an assistant you would like a beer.
"Let's say I say, I'd really like a beer, and at some point it's going to be, but why don't you try a Budweiser? Why don't you try a Coors Light?"
Today it would probably ask what you like, or walk you through a few styles. My worry is the day a brand name shows up first and you cannot tell why it got picked. That is my prediction. OpenAI has not announced anything like it, and I will keep separating the two all the way through this piece.
Follow the money: who pays, who gets paid, who eats the loss
I use the same rule on AI that I use on a real estate deal: when you want to understand why something behaves the way it does, follow the money.
Ask three questions. Who pays? Who gets paid? Who eats the loss when it goes wrong?
With a pure subscription, the answers are simple. You pay. The company gets paid. If the product disappoints you, the company eats the loss, because you cancel. The product works for you because you are the only customer in the room.
Advertising brings a second paying customer through the door. The advertiser pays, the platform gets paid, and you now share the room with someone whose money matters to the company. If a bad ad leads you to a bad purchase, the platform already collected its fee. The loss lands on you, and maybe on the better competitor who did not buy the spot.
"Now it's going to be more about who's paying the most for that ad space."
None of that makes advertising evil. Ads pay for a lot of free things people love, and a service used by 1.2 billion people a week has enormous bills. Somebody covers them. The only question worth asking is how the product changes once a second customer is paying too.
I also said on the LIVE that I expect the ad push to land hardest on people below the twenty-dollar-a-month plan, because the company has to make money from the people who are not paying. That is my guess about how these things usually shake out. Nothing in the announcement I am working from names which plans see the test, so treat my guess as a guess.
I watched search and social make the same trade
I have been in real estate since the late 1990s. I was alive and kicking when search engines started selling ads, and I watched social feeds do the same thing after that. I will not pretend to remember exact dates. I remember the pattern, because my business lived inside it.
From where I sat, it went like this. A free tool gets useful. Everybody shows up.
For a while, businesses get found for free because the results are about relevance. Then paid placement arrives, usually labeled and off to the side. Over the years, the paid part takes up more of the page, and the free part has to fight for room.
For a small business, that changed the math. In my world, the agents who had built their own websites and kept in touch with their own past clients kept getting calls. The ones who only rented attention had to keep paying rent, month after month, at a price someone else set. I watched good people get squeezed that way.
On the LIVE I called this possibly another Facebook moment, possibly another pay-per-click moment like Google. That comparison is my opinion. I am not saying ChatGPT will copy either company. I am saying that every time I have seen a free tool add a paying advertiser, the business model shaped the product over time, and a smart owner plans for that before it happens.
A labeled ad next to a picture is not a paid answer
Two very different things get lumped together when people argue about AI ads. Business owners and everyday users both need to keep them apart.
The first is what OpenAI actually announced. You ask for an image. While it generates, a clearly labeled ad appears in its own space, separate from your picture. You know it is an ad. You can ignore it the same way you ignore a banner across the top of a news site.
The second is a recommendation inside an answer. You ask for the best plumber near you, and the answer names a company because that company paid. No label, no separate box, just the voice you have learned to trust saying a name. That second thing is my concern. It is not what was announced, and OpenAI says ads do not influence answers.
The gap between those two is the whole ballgame. A labeled ad asks for your attention, and you decide what to do with it. A paid recommendation would borrow the tool's credibility, and you would never get the chance to discount it, because you would not know it was paid. One is advertising. The other would be a referral you did not know was bought.
My worry is a concern to watch, not a fact
I said on the LIVE that "ads change who the product works for." I believe that. In the same breath I admitted something just as important: I do not know how the ads will end up looking, whether off to the side or somewhere closer to the answers. Both things are true at once, and I want you holding both.
I can tell you my concern. I cannot tell you it has happened. The company says ads do not touch answers, and I have no evidence that says otherwise. So instead of treating a hunch like news, here is what I am watching for over the coming months:
- Ads moving out of image generation and into ordinary chat replies.
- Ad labels getting smaller, lighter in color or easier to scroll past.
- A business name showing up first in an answer with nothing marking it as sponsored.
- Answers to "best" and "near me" questions shifting after the test grows, with no change in the businesses themselves.
- Any rewrite of the public statement that ads do not influence answers.
If none of that shows up, good. I will say so on the show. If some of it does, you will have watched it happen with your eyes open instead of waking up one morning to a different product.
Users also hold more cards than they think. There are several capable assistants now, and switching costs you an afternoon. A company with 1.2 billion weekly users has a lot to lose if people decide the answers feel bought. That competition is the user's best protection, and I think the companies know it.
AI is a very good salesperson
"Is AI a good salesperson? Yes. Better than a human? Absolutely."
That is a strong thing for a guy who has sold houses for decades to say on camera. I said it because I have watched what these tools do in a conversation. Three traits explain most of it.
It is patient. It never gets tired of your fifth question at eleven at night, and it never sighs. It is personal. It remembers what you told it ten minutes ago and frames the answer around your own words, your budget, your worry.
It is always on. No bad days, no lunch break, no Monday mood. The best human salespeople spend whole careers trying to build those three habits, and the machine starts with all of them.
And it works on me. I am 57. I would have told you TikTok Shop was a dumb idea and I would never buy anything there. I have bought things off TikTok Shop, and I have not been unhappy with one yet. If a feed of short videos can turn a guy like me, a patient assistant that knows what I asked for an hour ago is a whole other level of persuasion.
None of that is a reason to fear the tool. A great salesperson can help you buy the right thing faster and skip the junk. It is a reason to know who that salesperson works for, which takes us right back to following the money.
Hide your credit card: slow down before you buy inside a chat
"Be careful. Hide your credit card. Don't give it up right away."
I remember when buying anything online sounded a little crazy. People would ask if you had bought anything online yet, and the very next question was fear. What if somebody steals my card number? What if they drain my account? I am not sure I even had a credit card back then.
Some people did get burned. It was not enough to stop anyone, and today paying online is as normal as paying at a register.
I expect buying from inside a chat to follow a similar path. Early caution is not paranoia. It is the price of being early to anything. The trick is to stay careful without refusing to learn the new thing.
So build a few habits now, while the stakes are small. When an assistant points you toward a product, open the store's own website in a separate window and read it there. Look for the ad label. Ask the assistant directly whether the suggestion is paid, and ask for three alternatives it would put next to it. For anything you would miss the money on, sleep on it one night.
Do not save your card inside a tool you are still figuring out. Read the permission screens before you tap through them. If an assistant ever offers to buy something on your behalf, decide ahead of time what it may do alone and what it must bring back to you first. I wrote a full piece on that boundary: AI can help, but you still need to be in control.
The ad decision a business owner faces this month
If you own a business, the first question is whether to test ChatGPT ads when you get the chance. Fair question. Start with what the test actually is: visual ads, during image generation, in the U.S., with an initial group of advertisers. Then ask whether your customers are the kind of people who make images in ChatGPT, and whether a labeled ad beside their picture would ever turn into a phone call.
If you do test, run it like any new channel. Pick one offer. Set a budget you can afford to lose completely.
Decide before you start how you will know if it worked, meaning calls, booked appointments or sales, not clicks. Then stop on the date you set and look at the numbers. My piece on turning an AI idea into a bounded business test walks through that discipline. I am not your accountant, and this is not financial advice. It is how I would think about a new ad channel with my own money.
The ad decision is the smaller one, though. The bigger shift I talked about on air is that many owners are now using ChatGPT itself as their marketing partner. They write their own ads, their own emails, their own posts. I said on air that some of those people might not need a marketer like me anymore. That change cuts across a lot of jobs, and I dig into the bigger economic question in AI replacement has a buyer problem.
I still think people care about the human being in the deal. I just would not bet my rent on that alone.
Make sure the assistants can find you without paying
Before you spend a dollar to stand next to somebody's image, find out what the assistants already say about you for free. This is the part most owners skip, and it is the part that keeps paying off whether ads grow or stay small.
Sit down with two or three assistants and ask them the questions your customers ask. Who does your kind of work in your town? What does your business do? What are your hours, and how do I reach you?
Read every answer slowly. Wrong hours, a service you stopped offering years ago, an old phone number, a competitor's name where yours should be: all of that is fixable, and you fix it at the source.
The source is your own website and your own public business information. Clear service pages written in plain English. A contact page that works. The same name, phone and address everywhere you are listed.
My guide on whether AI agents can find and understand your business covers what to publish and how to measure what happens next. Nobody can promise you a ranking inside an AI answer, and I will not. What you can control is whether the tool has accurate material to work from when it describes you.
If ads ever do drift closer to answers, the business with clean, accurate public information still starts in a better spot than the business that never checked. If they never drift at all, you still won, because customers who ask about you get the truth.
Own the relationship, track the source, ask the money question
The lesson I took from watching search and social is that rented attention can be repriced overnight. Owned relationships cannot. Here is what I would put in place this month, whatever you decide about ads:
- Keep your customer list in a system you control, with clear permission to contact each person.
- Keep your own website, email and text channels, so a change on someone else's platform cannot cut you off from your people.
- Ask every new lead how they found you, and write down the exact answer, including "ChatGPT told me about you."
- Tag each lead by source so you can see whether AI-referred leads grow or shrink as the ad test expands.
- Ask every AI tool you use how it makes money: subscription, ads, referral fees or something else.
- Reread a tool's terms any time it changes its pricing or announces advertising.
That money question is the one I want you to get comfortable asking out loud. It is not rude. A tool you pay for, a tool that runs ads and a tool that earns a fee when you buy something each have a different reason to point you somewhere. You do not have to avoid any of them. You just have to know which one you are using before you let it steer a decision.
Lead tracking matters more than it sounds. If you do not know where your calls come from today, you will never notice when an AI assistant starts sending you fewer of them, or more. A simple note on every new contact is enough to start. Six months from now, that note will tell you more about AI and your business than any headline will.
How I build my own tools: Clara tells you what she is
I hold my own tools to the standard I am asking of the big companies. On SantaClaritaOpenHouses.com, the assistant is named Clara. The first thing she does is tell you she is an AI. Nobody has to guess, and nobody gets fooled into thinking they are texting a person.
She also gives the market update with a date and time stamp. Market numbers in Santa Clarita move through the day, and a figure without a time on it is a figure you cannot judge. You can read how she is set up on the Clara reference page.
The same principle runs my real estate business. I am a Sellers Only Agent™, which means I work for the seller, and I say so up front. Telling people who you work for is not a slogan. It is the same disclosure I want from every AI tool that recommends anything to anyone.
That is the contrast I want you to carry out of this piece. A tool that tells you what it is, who it works for and how fresh its information is earns trust the right way. A tool that hides any of those three is asking you to trust the tone. Ads in ChatGPT may turn out fine. Either way, you will be ready, because you will be asking the right questions.
Where to go next
Can AI agents find and understand your business? Start here if you have never checked what the assistants say about you.
AI can help. You still need to be in control. Read this before you let any assistant act or spend on your behalf.
An AI idea is a starting point. Use it to turn an ad test into a small, measured experiment.
AI replacement has a buyer problem. The bigger economic question behind owners doing their own marketing with AI.
Book time with me. Bring one question about what the assistants say about your business, and we will work out what to fix first.