Location Targeting Inside ChatGPT Ads: Country, Region, City, DMA and Postal Code

Last verified: 2 October 2026 | Version: 2.0 | Next scheduled review: 2 November 2026

You can target a country everywhere. Below that, OpenAI's campaign documentation now says you can target "supported locations such as states or regions, cities, markets, and postal codes where available", and that availability varies by country. It does not publish which countries get which levels.

The first version of this page, written on 12 September, said you could not target a city at all, because OpenAI's documentation then named only state, DMA and postal code in the United States. The documentation has since added cities. So the honest answer to "can I target a city?" is now: search for it in the campaign picker, and if it resolves as a city, yes. If it only resolves as a DMA, the trap described below still applies.

First, a distinction that gets collapsed

Two different questions both get called location targeting, and they have different answers.

Can ads serve in this country at all? That is a serving market question, decided by OpenAI, and it is fixed. It is also separate from where your business is registered, which determines whether you can hold an advertiser account in the first place. Our eligibility post owns both of those.

Within a market where ads do serve, where do I want mine to appear? That is targeting, and it is the campaign setting described here. Nothing you do in this setting extends the platform's serving markets. You are dividing a permitted area, never adding to one.

If you are in a market that does not serve, no targeting configuration helps and there is no workaround worth attempting.

The four levels, and what they are available for

Two constraints sit around that table.

Subnational targeting outside the United States is still uncertain. OpenAI's documentation confirms country targeting everywhere and lists finer levels "where available" without naming markets. Our 8 August count of OpenAI's geotargets file found prefectures in Japan, cities and provinces in South Korea, regions in New Zealand, provinces in Canada, states in Australia and nations in the United Kingdom, and nothing below country level anywhere else; the country pages carry those findings market by market. OpenAI's advice is to search the location in the campaign picker, which is a polite way of saying the list is not published.

There is a second constraint that has nothing to do with granularity. OpenAI says some new self-serve accounts can initially advertise only in their home country, and choosing other countries does not override it. Location targeting cannot get a new account past that.

And campaigns built from product feeds are country-level only, with exclusions, though existing campaigns may retain previously saved subnational settings. If you plan to move to feeds later, the geography you built around may not survive the move.

The DMA trap

This is the failure that costs money, and it is structural rather than accidental.

A designated market area is a Nielsen construct built for television. It groups counties by which local broadcast signals they receive, which means it was drawn around transmitter reach rather than around anything a marketer cares about. Where the picker does not resolve a city, the DMA is the closest unit to one, so "I want Austin" becomes "I have selected the Austin DMA". Even where a city level now exists, it is easy to pick the market by habit.

The two are not the same thing. Some examples of the gap, using standard Nielsen DMA definitions:

• The New York DMA extends well beyond New York City into northern New Jersey, parts of Connecticut and parts of Pennsylvania.
• The Austin DMA covers surrounding Texas counties, not the city limits.
• The Washington DC DMA spans the District, northern Virginia and much of suburban Maryland.

For a B2B advertiser this is usually tolerable, because the buyers you want are distributed across a metro area anyway and the extra reach is not waste. For a local service business with a genuine travel radius, it is the difference between a qualified enquiry and a call you cannot service. That case is covered in our local services post.

The practical consequence: if your targeting rationale contains the word "city", check whether the picker offers the city itself before reaching for the DMA. If it does not, ZIP codes are the right instrument. Build the ZIP list from the geography you actually serve rather than selecting the metro and hoping.

Where geography stops being a targeting tool

A point that is easy to miss on this channel specifically.

On search, geography is partly a proxy for intent, because people type their location into queries. Here there are no queries. Location is a hard filter applied to the user, and it carries none of the intent signal it carries elsewhere.

So location narrows, and narrowing has the usual cost on a channel where underdelivery is the common failure. Restricting a campaign to three DMAs at a 25 USD daily minimum can leave you with a campaign that does not spend, and no way to tell whether the cause was the geography, the hints or the bid, because none of those are reported separately.

The sequencing that avoids this: start at country level, confirm the campaign can spend, then narrow. Starting narrow and widening is the same number of steps but it starts with your least diagnosable state.

The reckoning: this contradicts the tidy-targeting instinct

Experienced buyers tighten geography early, because on Google and Meta it is cheap insurance against waste. That instinct imports badly here.

Tightening geography on a channel with no search terms report, no hint-level reporting and no location breakdown in the reports means adding a variable you cannot see the effect of. Ads Manager reports at campaign, ad group and ad level. There is no geographic dimension in reporting, so unlike platform, which Insights can now segment, the only way to compare two geographies is to run two campaigns and split the budget.

So the honest position is that most B2B advertisers should run country level for the first month, not because broad is better but because it is the only configuration where an underdelivering campaign has fewer possible causes. Narrow when you have a specific reason: a licence that stops at a state line, a sales team that only covers two regions, a genuine service radius. Do not narrow because it feels disciplined.

What we cannot tell you

• Which markets support city and other subnational targeting, and at what granularity. Not published. OpenAI directs advertisers to the campaign location picker instead of a list.
• How location performance compares. No geographic dimension exists in reporting, so no within-campaign comparison is possible.
• How a user's location is determined. OpenAI does not document the signal used, its precision, or how VPN and travel are handled.
• Whether DMA boundaries used by OpenAI match current Nielsen definitions exactly. Not stated in the documentation.
• What a narrow geography costs in delivery. Not measurable without a geographic reporting dimension.

Quick answers

Can I target a city in ChatGPT Ads? Where available. OpenAI's documentation now lists cities among supported location types, without saying which markets have them. Search the city in the campaign picker. If only a DMA resolves, remember it covers several counties; use ZIP codes for genuine precision.

What location levels does ChatGPT Ads support? Country everywhere, plus states or regions, cities, markets and postal codes where available. The United States has state, DMA and ZIP code. Availability elsewhere is not published.

What is a DMA? A designated market area, a Nielsen media market grouping counties by broadcast signal reach. It was built for television, not for marketing radii, which is why it rarely matches a city.

Can location targeting let me advertise in a country where ChatGPT Ads does not serve? No. Targeting divides a permitted area. It never extends the platform's serving markets, and account eligibility follows your country of business registration separately.

Can I see which locations performed best? No. Reporting runs at campaign, ad group and ad level with no geographic dimension, so comparing two geographies requires two campaigns and a split budget.

Should I start broad or narrow? Broad. Country level for the first month leaves an underdelivering campaign with fewer possible causes, then narrow when you have a licensing, coverage or service-radius reason.

Sources

Related reading

• ChatGPT Ads by country: minimum budgets and rules
• ChatGPT Ads Eligibility: Which Countries and Which Businesses
• ChatGPT Ads Eligibility Checker
• ChatGPT Ads for Local Services
• Platform Targeting on ChatGPT Ads: the Setting Nobody Configures
• How to Structure a ChatGPT Ads Account When There Are No Keywords

Changelog

2 October 2026, v2.0. OpenAI's documentation now lists cities among supported location levels "where available". Withdrew the claim that city targeting does not exist, reframed the DMA trap as the fallback when no city resolves, added the home-country restriction and our August per-country count, and corrected the platform reporting comparison.

12 September 2026, v1.0. First publication. Documents the absent city level and the DMA substitution, and separates serving-market eligibility from targeting inside a permitted market.

Field kit

Tools

• Eligibility Checker
• Budget Calculator
• OAI-AdsBot Checker
• Context Hints Generator
• Buy Scorecard

Site

• ChatGPT Ads
• Google AI Mode Ads
• Methodology
• Glossary
• Contact

OpenAI (primary)

• ChatGPT Ads Help Center
• Ads in ChatGPT: The Basics
• Measure Results
• Ad Policies
• Ads Manager
• OAI-AdsBot verification