ChatGPT Ads for Local Services: The Approved Category Nobody Writes For
By Ansh Khandelwal 9 min read
In brief
Local services is a permitted category. Sub-national targeting exists at US postal-code level only — which decides whether the channel works locally.
Last verified 2 October 2026. Version 1.1. Next scheduled review 2 November 2026.
Read OpenAI's permitted advertising categories and then read the guides written about them, and the mismatch is hard to unsee.
Permitted: household and consumer goods, local services, travel and experiences, digital products, education.
What the category writes about: ChatGPT ads for B2B SaaS. Dozens of guides, several agencies specialising in it, an entire content cluster.
B2B SaaS is not on the permitted list. Local services is, and almost nobody is writing for the plumber, the clinic, the driving school, the moving company or the wedding photographer, all of whom are explicitly allowed to advertise here.
We noticed this the same way anyone would. The keyword export we pulled on 8 August returned 25 advertiser-side queries, and one of them was how to advertise bounce house events on chatgpt, on TikTok. A local operator asking a question that no guide in this market is written to answer.
Why this fits better than it looks
The audience exclusions work in your favour. OpenAI does not show ads to users on Plus, Pro or any Business plan. That exclusion is what makes the B2B case structurally weak, because enterprise buyers sit on work-provisioned accounts. It does the opposite for a local business, whose customer is a consumer on a Free or Go plan asking a question about their own life.
The query shape suits a conversation. Somebody choosing a local service is not usually running one search. They are working through a decision: whether they need the service at all, what it should cost, what to ask, how to tell a good provider from a bad one. That is a conversation, and it is a longer window than a single result page.
The competition is not here. The agencies selling ChatGPT ads are selling to B2B, because that is who pays retainers. The auction for local service categories is thinner than it will be.
The constraint that decides everything
Sub-national targeting is deepest, and best documented, in the United States.
Update, 2 October 2026. OpenAI's campaign documentation now says campaigns can target "states or regions, cities, markets, and postal codes where available", with availability varying by country, and still publishes no per-country list. The table below is our 8 August count of the geotargets file. It is the most recent public enumeration we know of, but the live campaign picker may now offer finer levels, including cities, in some markets. Search your service area in the picker before deciding.
We pulled OpenAI's full targeting catalogue on 8 August 2026 and counted it: 42,110 rows, of which 41,488 are United States postal codes.
| Country | What you can target below country level |
|---|---|
| United States | 57 regions, 210 TV regions, 41,488 postal codes |
| Japan | 47 prefectures |
| South Korea | 17 cities and provinces |
| New Zealand | 17 regions |
| Canada | 13 provinces and territories |
| Australia | 8 states and territories |
| United Kingdom | 4 nations |
| Everywhere else | Nothing, as of 8 August |
For a local business this is the whole decision.
A plumber in Denver can target postal codes and buy something close to a real service area. A plumber in Manchester can target "England," which is 57 million people to sell a two-hour callout radius. A plumber in Toronto gets Ontario. A plumber in Auckland gets the Auckland region, which is actually workable.
On the August file, this channel was viable for US local businesses, marginal for New Zealand, Japan and Korea depending on how your service area maps to the available units, and close to unusable in the United Kingdom, Canada and Australia unless your service is genuinely national.
Read plainly: for a local business, the qualification test is whether the campaign picker resolves your actual service area. In the United States it does, down to postal code. Everywhere else, including the 50-plus markets that opened since August, check it in the picker before you build anything. If the smallest unit it offers is far bigger than your callout radius, the channel is not for you yet.
The inventory question a local business has to ask
Being in a permitted category does not mean your conversations carry ads.
No ads run beside conversations about personal health, mental health or politics, or in what OpenAI's policy calls sensitive user journeys. For most local services this is irrelevant. For some it removes the moment of purchase entirely.
A dentist, a physiotherapist, a therapist's practice, a home care agency: these are local services by any ordinary definition, and the conversations that precede hiring them are frequently conversations about personal health. Health services also sit in the case-by-case review tier with proof of licensure, and that route is a United States carve-out, so a clinic outside the US is in a different position again.
The test to run before budgeting: write down the actual conversation a customer has before hiring you. If it reads like a health conversation or a personal crisis, your inventory is thinner than your category suggests.
What a sensible first test looks like
Check the postal code map first. If your market is outside the United States and your service area is smaller than a province or a nation, stop here. That is not pessimism, it is the targeting file.
Budget for interpretable volume, not the floor. The minimum is 25 USD a day per campaign, and a two-week test at the floor produces single-digit conversions for most local businesses. For a service with a handful of jobs a month, that cannot distinguish a working channel from a dead one.
Instrument the phone, not just the form. Conversion measurement exists in Ads Manager Beta and UTM parameters persist on ad clicks, so web conversions are trackable. Phone calls are not, and local services convert by phone more than by form. If your tracking only counts form submissions, you will measure the wrong half of your own business.
Check the crawler before you spend anything. Landing pages that block OpenAI user agents are a listed cause of ad rejection. Local business sites are frequently on managed platforms with CDN rules the owner never configured and cannot see.
What we cannot tell you
Whether local services perform. No published data exists for this category specifically, and OpenAI states no benchmarks exist across advertisers, industries or campaign types.
Which markets now have finer targeting. OpenAI's documentation now names cities among the supported levels "where available", without saying where. Until it publishes a list, the picker is the only source.
How the assistant chooses which local provider to surface organically. Relevant, because the ad sits beside an answer that is already recommending someone. Not documented.
Anything from running these. We do not operate campaigns. This is documentation, our own measurement of the targeting file, and reasoning from both.
Quick answers
Can local businesses advertise on ChatGPT? Yes. Local services is one of the permitted categories under OpenAI's ad policies v1.6, alongside household and consumer goods, travel and entertainment, digital products and education. Local health, legal and financial providers are a different matter: those categories are case by case in the US and generally prohibited elsewhere.
Can I target a city or postal code with ChatGPT ads? In the United States, postal codes yes, across 41,488 of them in our August count. OpenAI now lists cities among supported location types "where available" without naming the markets. Our August count 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 country level everywhere else. Search your area in the campaign picker to see what it offers today.
Are ChatGPT ads worth it for a small local business? It depends almost entirely on geography. A US local business can target a realistic service area. A local business in the UK, Canada or Australia can currently only target a region far larger than it serves, which makes the spend inefficient regardless of creative.
Can a dental or medical clinic advertise on ChatGPT? Health advertisers fall into the case-by-case review tier with proof of licensure, and that route applies inside the United States. Health services advertising outside the United States is generally prohibited. Separately, no ads run beside conversations about personal health.
Why is every ChatGPT ads guide written for B2B SaaS? Because that is who buys agency retainers. B2B software does not appear on OpenAI's permitted category list, while local services does.
How do I track phone calls from ChatGPT ads? Not through the platform directly. Ads Manager Beta supports conversion measurement and UTM parameters persist on ad clicks, which covers web conversions. Call tracking has to be handled on your own landing pages with your own call-tracking numbers.
Sources
| Claim | Source | Tier |
|---|---|---|
| Permitted categories include local services | OpenAI ad policies v1.3, 15 July 2026 | Confirmed, primary |
| Financial and health case-by-case with licensure; outside the US generally prohibited | Same | Confirmed, primary |
| No ads beside personal health, mental health or political conversations, or sensitive user journeys | Same | Confirmed, primary |
| Ads not shown to Plus, Pro or Business plan users, or predicted under-18s | OpenAI Help Center, updated 30 July 2026 | Confirmed, primary |
| Targeting catalogue: 42,110 rows, 41,488 US postal codes, 249 countries selectable | Our pull of developers.openai.com/ads/openai-geotargets.csv, 8 Aug 2026 |
Confirmed, our measurement |
| Sub-national granularity by market | Same | Confirmed, our measurement |
| Location types "where available", including cities | OpenAI Help Center, Create Campaigns for ChatGPT Ads, 2 Oct 2026 | Confirmed, primary |
| Minimum daily spend 25 USD per campaign | OpenAI advertiser documentation | Confirmed, primary |
| Advertisers receive aggregated performance data, no conversion attribution | OpenAI documentation and user communication | Confirmed and reported |
| Landing pages blocking OpenAI user agents are a listed rejection cause | OpenAI Help Center, Troubleshooting | Confirmed, primary |
| No benchmarks across advertisers, industries or campaign types | OpenAI advertiser FAQ | Confirmed, primary |
| Local-intent query observed on TikTok search | Our keyword export, 8 Aug 2026 | Confirmed, our measurement |
Related reading
- Can my industry advertise on ChatGPT? — the three category tiers and the review gap
- Who should not buy ChatGPT ads yet — including the test-size problem
- What a readable ChatGPT ads test actually costs
Changelog
2 October 2026. v1.1 added OpenAI's new documentation of city and other sub-national targeting "where available", reframed the August catalogue count as historical, and updated the category reference to policy v1.6.
15 August 2026. v1.0 published.
Field kit
Tools
Site
OpenAI (primary)
Share this
Ansh works across GEO strategy, B2B research, and execution. At InPromptAds, he translates new AI advertising products into clear operating advice, tests, and measurement questions for marketing teams.
