ChatGPT Ads opened to all US businesses. Wait six months.

The headline reads like permission. Read the fine print first.

OpenAI opened the ChatGPT Ads Manager to every US business on May 5 2026. There is no minimum spend, no agency, and no waitlist. Anyone with a credit card can sign up tonight, write a short description of when the ad should show, and start paying for clicks by morning. The launch was covered in every paid-search newsletter this week. Plenty of operators will register an account by Wednesday.

Hold off if you serve one city. The reason is buried in the targeting settings. The only option for where the ad runs is the United States. State, city, ZIP, and radius filters remain absent from the dashboard at this stage of the beta. A hundred-dollar daily budget on these ads buys you clicks from every state in the country, including the 49 outside your service area. The math will work later in 2026, when OpenAI ships city-level targeting. Until then, the budget belongs somewhere else.

What the launch actually shipped

The pilot version of ChatGPT Ads has been running since fall 2025 with a $50,000 minimum and an approval queue. That minimum is gone. On May 5, OpenAI rolled out a self-serve dashboard, which means you sign up and run it yourself with no salesperson. You upload your ad, set a daily budget, and pick how much you are willing to pay per click. The official OpenAI post titled "New ways to buy ChatGPT ads" and a Search Engine Land walkthrough by Greg Sterling both went up the same morning. Both confirm the same three facts.

First, any US business can register and run a campaign. Second, eligible categories include household goods, local services, travel, entertainment, digital products, and education. Third, the geographic setting (the option to pick which state or city the ads run in) is set to the whole United States and nothing smaller. State, DMA, city, and radius targeting are not in the dashboard yet. OpenAI has said the smaller settings are on the roadmap. They have not given a date.

The second piece worth understanding is the way the ads actually decide where to show. ChatGPT does not let you pick keywords the way Google does. Instead, you describe in plain English the kind of conversation the ad should sit next to, and OpenAI decides where it shows. Your description might say "the user is asking how to fix a slow drain in an old house." The AI matches that note to live chats and decides on the spot whether your ad belongs there.

That part is interesting. It is also why the lack of a city setting hurts more than it would on a regular paid-search platform. The plain-English description can be the best one ever written about slow drains in old houses, but the ad still shows up in chats from anywhere in the country. A plumber who only serves one ZIP code is paying for every click that comes in, including the ones from a thousand miles away.

Where the money goes if you turn ads on this week

Picture a four-room medspa in one ZIP code. Set a $100 daily budget. Set the targeting to the United States, since that is the only option. Write a tight description that says the ad belongs near chats about glow-up skincare for women in their thirties. The AI does its job and shows the ad in matching chats across the country.

The medspa pays for every click. The clicks come from chats in 50 states. Maybe one out of every fifty clicks comes from a person who lives within a 30-minute drive of the medspa. The other 49 clicks are pure waste. After a week of running, the medspa has paid for 800 clicks and seven of them are people who could ever book an appointment.

The brands that did well in the Criteo pilot ship to every doorstep in the country, which is why the country-wide setting works for them. A practice with one waiting room and a parking lot is the opposite case. Until the dashboard ships a city or radius setting, the entire targeting model is built for advertisers who can sell to anyone.

What the Criteo pilot numbers do and do not tell you

The pilot ad data is real. Brands running through Criteo's integration with OpenAI saw conversion rates close to two times higher than regular paid search, per PPC.land coverage in May 2026. That is a solid lift. It is also from a pilot dataset that skews household goods, consumer electronics, lifestyle and wellness, and home and garden. Not one single-location service business has published a case study yet that says the math works for a one-zip-code business.

Compare that with what the same AI traffic does on the organic side. Visitors who arrive at your website from a Perplexity citation book at roughly eleven times the rate of a regular Google visitor, per Similarweb and Authority Tech analyses of Q1 2026 traffic. The reason is intent. A person who reads a ChatGPT answer about your service and then clicks the link in the answer has already done most of the homework on a vendor, and they clicked because they want to talk to that vendor. A click on a paid ChatGPT ad is a different animal. There is no vendor research behind it yet.

What to do with the money this quarter

The work that moves the phone for a one-location operator right now is on the organic side. Customers ask ChatGPT, Perplexity, Gemini, and Google's AI mode for vendors every day, and the answer is built from what those engines find on your website and inside your reviews. Each engine looks for plain readable text that answers the exact question a customer typed. Many service-business websites still publish marketing lines about quality and experience, so the model skips that page and quotes a competitor whose page actually has the answer.

Three actions, in order of payoff for the median service business. Write a short answer page for each of the five questions your front desk hears every morning before a customer books. Put a price (a range is fine) on every service you sell. Reply to your last 20 reviews in plain sentences that name the service and the question the customer asked about. The combined work is one or two afternoons.

One more piece, because the engines do not agree. ChatGPT and Perplexity share only 11 percent of the websites they cite for the same question, per an Authority Tech audit of 12,000 commercial queries in Q1 2026. Showing up on one engine does not mean you show up on the others. Run the same five questions through ChatGPT, Perplexity, and Google's AI mode separately. Write down what each one says. The gaps tell you what to write next.

If this fits your business, run a free audit at vyzz.io. Seven AI engines included, with the audit skipping the call, the pitch, and the credit card step. You will see what each one says about your business in about 90 seconds, and you will know whether to revisit the ads question in November.