How Do Customers Use ChatGPT to Find Local Services?

The first question is usually simple

Most of the early questions are simple and keyword-based. Local searches in ChatGPT do not always begin as long conversations. A customer may type "dentist near Boston," "emergency plumber in Austin" or "wheelchair-accessible restaurant nearby." The first request can be as short as the search it replaces.

Sagapixel observed people in the United States and Canada while they looked for dentists, dermatologists, chiropractors, cosmetic clinics and surgeons. Its study says 75 percent of the sessions contained at least one keyword-style search, and 45 percent ended after one prompt. The study page reports 177 sessions; the downloadable prompt sheet contains 176 data rows. That mismatch is one reason to treat the results as useful field evidence rather than a universal benchmark.

OpenAI's own documentation helps explain why a short question can still produce a local answer. ChatGPT Search may rewrite a request into one or more targeted searches. It can use an approximate location based on an internet connection, and a more precise device location only when the user turns that option on. A customer does not have to state every detail in one sentence for the system to add context.

This does not mean the first answer settles the decision. Sagapixel found that people considered several businesses on average. Some accepted the first list, some asked for more choices and some began checking the differences that would affect an appointment.

The comparison gets more specific

The follow-up questions show what a customer still needs to know. AnswerPrism coded the follow-up field in all 176 rows of the public prompt sheet. Twenty-six rows mentioned reviews, ratings or reputation. Nineteen mentioned price, affordability, discounts or insurance. Thirty asked for a website, address, directions, phone number, booking, scheduling or hours.

The groups overlap. Fifty-two rows contained at least one of those three kinds of question, 16 contained two or more, and seven contained all three. The count is descriptive, not a claim about how every customer searches. The source covers healthcare and aesthetics, and some rows record an option that ChatGPT offered and the participant accepted.

Still, the examples are recognizable outside a clinic. One person wanted a dentist that took a named insurance plan and had afternoon or evening hours. Others asked for a price range, customer reviews, a provider's website or the next available appointment. The same pattern applies outside clinics: fit, cost, proof and a practical route forward.

A roofer may be compared on emergency availability and the kind of roof repaired. A restaurant may be compared on dietary fit, price and reservations. A lawyer may be compared on jurisdiction, practice area and consultation terms. The useful detail is the one that changes whether this business can handle this customer's need.

Publish the facts that survive the follow-up

The first page a customer sees should answer the service and location question. The next useful details belong close enough that a person can keep making the decision without decoding the company.

Google gives eligible Business Profiles places for many of these facts. Certain businesses can add services with descriptions and prices, factual attributes such as accessibility, customer-facing hours and links for booking or ordering. The available fields vary by category and country. The website should carry the fuller explanation, especially when price, eligibility, timing or scope needs context.

The pricing guide explains when to use an exact price, a range or a quote process. Whatever format the business chooses, it should answer the financial question honestly enough for a customer to decide whether making contact is sensible.

Do not treat a completed profile field as proof that ChatGPT will choose the business. OpenAI does not publish a rule that rewards a particular Google field, and Google does not say every field applies to every business. Publish the information because a customer needs it and because it gives the public record a clearer answer.

• Name the specific service, meaningful limits and the area where it is available.
• Explain price with an exact figure, a defensible range or a clear quote process.
• State ordinary hours, appointment rules, response windows and after-hours limits.
• Put inspectable proof near the claim it supports, such as a licence, named credential, comparable work or a review about the same kind of job.
• Give the customer a working next step: call, book, order, request a quote or get directions.

One answer is not the whole decision

ChatGPT can provide citations and a source panel when it searches the web. OpenAI also warns that its search results and citations may be incomplete, outdated or wrong. The existence of a source link suggests that the information may originate from a particular source, but it does not necessarily explain why the particular business appears in the answer. A careful customer may open a source, search the business name, read reviews elsewhere or call without visiting the website at all.

A business can appear in the first answer and lose the customer when the website hides the price process or the booking link fails. It can be absent from one answer and still be found when the customer asks for a more specific service. A mention does not show whether the person trusted the claim or completed the next step.

Consistency is more than just making sure that the exact text is identical. A short profile can state that evening appointments are available. The relevant service page can explain which days, whether new customers qualify and how far ahead to book. The booking tool should then show the same real availability.

Test the decision, not the business name

A useful check begins with a request a real customer might make. Start with the service and market, then add one constraint that changes who fits. Use a fresh ChatGPT Search conversation and save the answer and visible sources.

Follow the path a buyer might take:

The test is not a performance staged for the company name. It should reveal whether the business remains a credible option as the question becomes more specific. A wrong service area, missing price process or dead booking link is a concrete editing job. One unfavorable list is not evidence that the entire market presence has failed.

1. Ask for the service in the city, neighborhood or postal code.
2. Add one real constraint such as timing, budget, property type, insurance or accessibility.
3. Ask why the named businesses fit that requirement.
4. Ask for the price information, reviews or proof needed to compare them.
5. Ask for the website, phone number, directions or booking route.
6. Open the visible sources and check every important fact before acting.

Record how the customer made contact

The final handoff may not leave a clean trail in analytics. A person can see the business in ChatGPT, search the name in Google and arrive as organic traffic. Another may call directly from the information in the answer.

Ask new customers how they found the business and preserve their wording. Compare that intake record with identifiable AI referral visits instead of forcing both into one attribution number. The measurement guide shows how to combine the GA4 record with what customers tell staff.

Track where customers stop, which facts they still ask staff to explain and which page or profile should answer them next time.