ChatGPT Ads for Local Services: 90 Million Weekly Requests and a Gap Worth Knowing

OpenAI's September 2026 stat pack includes one slide it labels differently from the rest: local services, over 90 million ChatGPT messages a week categorised as requests for nearby businesses and service providers, from an "internal LiDAR analysis" flagged as an exploratory estimate. Next to retail's 1.5 billion it is small, and next to Google's local query volume it is smaller still. That gap is the most useful thing on the slide.

What 90 million is, and is not

It is a weekly average of global messages, across 6 to 26 July 2026, in which the person asked ChatGPT for a nearby business or provider. It is the only category volume in the pack that OpenAI calls exploratory, and its footnote is the strictest: "messages, not users, ad inventory or verified purchase intent." It covers every country, and ads serve in a subset. It includes the person asking for a plumber today and the person asking which neighbourhood in Lisbon has good nurseries, which are very different buying moments.

Why local is where ChatGPT and Google differ most

On every other category OpenAI published, ChatGPT demand is within an order of magnitude of what a brand would expect from search. Local is the exception. Our reach calculator, which sizes ChatGPT demand for a situation against Google search volume for the same intent, produces its widest gap on local hints: OpenAI's own 90 million against retail's 1.5 billion is a ratio of about 0.06, and that ratio surfaces directly when you size a "dentist near me" style situation. The reason is structural. "Near me" behaviour lives in Maps and in the search box, with location attached automatically; ChatGPT has to be told where you are, and most people do not tell it for a quick errand.

What people do use ChatGPT for locally is the considered version of the same question: which type of provider do I need, what should it cost, what should I ask, is it worth travelling for. That is the discovery stage, and it is where a local business with a real point of view can appear before the person opens Maps.

Targeting, and the European catch

In the United States, ChatGPT Ads can target states, cities, DMAs and postal codes. In Europe, as of September 2026, campaigns target at country level. A Portuguese clinic cannot restrict delivery to Porto; it writes Porto into the context hints ("looking for a physiotherapist in Porto after a knee operation") and into the creative, and lets relevance do the geographic filtering. That works because the auction is relevance-weighted: an ad that names the city will not be shown against a Lisbon conversation often, and when it is, it will not be clicked, and the platform learns. It also means a single-location business pays the country-level minimum daily budget, roughly 650 euros a month, for a city's worth of demand.

When it makes sense

The honest answer for a single-location European business in 2026 is usually to spend the money on being recommended organically first: the same prompts, answered by the business's own pages, with no auction and no minimum. The channel starts making sense for a local business in three cases. A multi-location brand, where the country-level buy is the footprint anyway. A service people research for weeks before choosing, such as private schools, clinics, home renovation or wedding venues, where the discovery conversation happens on ChatGPT and the shortlist is written there. And a United States location, where sub-country targeting turns the channel into a genuine local buy. In each case the tracking sequence is the same as everywhere else: pixel and Conversions API, five conversions on the enquiry, then Conversion Optimization. The eligibility checker is the first thirty seconds, and the reach calculator will show you the local gap for your own situation before you spend anything.