How Small Service Businesses Get Found in AI Search

The short version

Over the 28 days ending September 17, 2026, Google Analytics credited ChatGPT with 24 sessions on the site of a music-lesson business I run. Fifteen were engaged sessions. In the latest seven days, ChatGPT sent eight sessions and six were engaged. The 28 days before that window had five total.

And in the last week, four booked students told me they found the business through ChatGPT.

What the data actually shows

The figures come from the site's Google Analytics property, with data through September 17, 2026, pulled on September 18. I wrote down the windows and pull dates so I can compare later.

Weekly ChatGPT sessions over the last eight weeks, oldest first: 0, 0, 2, 3, 7, 9, 0, 8. The last two days were still settling when I pulled it, so that newest number can move.

Across the full 56 days, the top landing pages were the homepage (10 sessions), a Los Angeles songwriting lesson page (3), a Los Angeles music production page (2), and an online piano lesson page (2). People landed on pages about specific services, not just the front door.

Two limits. Consent settings, privacy tools, and browser blockers all suppress analytics, so the real number of AI-assisted visits is higher than what got recorded. And an “engaged session” only means the visit lasted past a few seconds, hit a second page, or fired a tracked action. It doesn't mean anyone booked, paid, or became a customer.

What changed before the referrals

I can't prove which change produced these visits, and I'd be skeptical of anyone who says they can. What I can give you is the order I did the work in, because the referral trend followed it.

1. Made the pages reachable and indexable

Every service page loads, shows up in the internal navigation, sits in the sitemap, and doesn't accidentally block crawlers. Unglamorous, and it comes first. If a page can't be fetched, nothing can summarize it.

2. Gave every real service its own page

The business offers specific things in a specific area, so each one gets a page that says what it is, who it's for, what it costs or how pricing works, and how to start. No pile of near-identical pages with the city name swapped out. Those just compete with each other and give a reader nothing to choose between.

3. Made the business identity identical everywhere

The business name, service area, phone number, email, website, and credentials read the same on the site, the map listings, the professional directories, and the social profiles. When those disagree, every system reading them has to guess which version is real. Some guess wrong.

4. Showed proof a stranger can check

Credentials, years of practice, the specific instruments and software, published policies, and real examples. Reviews from real customers, linked back to the platform they came from. Nothing invented. Nothing that reads like a copywriter wrote a testimonial on somebody's behalf.

5. Wrote for the questions customers actually ask

The pages answer in the words people actually search with. What the service is, who it's for, what a first session looks like, what it costs, and whether they need to prepare anything. The direct answer sits near the top in plain language, before the long explanation.

6. Started measuring and asking

Referral sources get tracked in analytics, and every new customer gets asked how they found the business. So far, asking is the only method that has produced a firm answer. Last week it produced four.

The measurement most businesses skip

Analytics told me visits came from ChatGPT. It never told me those visits mattered. The proof that this brings in customers is people saying so, and last week four booked students said exactly that.

So I keep two separate records:

• Attributed sessions: visits where the referrer is an AI assistant. Good for trend, direction, and seeing which pages pull interest.
• Named discovery: a customer telling you, in their own words, how they found you. Small numbers, high value, and the only one that ties to revenue. Four of these came in a single week. That's a small sample, so I won't call it a rate.

Keep them apart. They measure different things, and adding them together gives you a number that means nothing.

There's a third, weaker signal. I keep a short fixed list of questions a prospective customer would ask and run them now and then in a fresh chat to see whether the business shows up at all. In one recent pass it was mentioned in four of the seven discovery questions, and three of the seven gave an answer that linked to the site. That sample is too small to quote as a percentage. It's a direction to watch.

What this does not prove

• It does not prove causation. Several changes landed in the same stretch, and answer systems change without notice or documentation.
• It is not a rate you can copy. One small business, one niche, one city, one reporting window.
• The student reports are self-reports. Four people told me they found the business through ChatGPT. That's what they said, not a tracked conversion, and I can't know how many would have found it some other way. Four is also a floor. Not everyone mentions how they found a business, and nobody counts the visitors who looked and left.
• It is not a promise. No search engine or assistant owes any business a recommendation. Clear structure, accurate details, and real proof raise the odds of being understood correctly. They don't buy placement.
• It is not a position you hold. A mention today isn't a ranking. The next model update can change it.

How to apply this to a service business

Here's the order I'd do it in again:

1. Write down what is actually true. Real services, real service area, real credentials, real prices or price ranges. Everything else on this list depends on it.
2. Inventory the pages you already have. One page per real service. Merge or delete the thin ones that only swap a city name.
3. Standardize the facts. Pick one canonical set of details, then fix your own listings and profiles to match before you create anything new.
4. Rewrite the two or three pages closest to a buying decision. Answer in the first paragraph, then back it up with details, limits, and what happens next.
5. Add proof. Credentials, examples, policies, and customer reviews hosted somewhere they can be checked.
6. Check the boring technical things. Pages load, mobile works, crawlers aren't blocked, the sitemap is current, and old URLs redirect instead of breaking.
7. Set up measurement and ask people. Track assistant referrals, and write down how every new customer says they found you.
8. Review on a schedule. Pages, listings, and answers all drift. The check only works if it repeats.

Three expensive shortcuts

• Location pages at scale. Forty pages, forty city names, one body of text. Readers get nothing, and the whole site looks manufactured.
• Mass-generated articles. Volume with no first-hand detail reads as filler, to people and to the systems summarizing pages. One good page beats ten thin ones.
• Treating markup as the strategy. Structured data helps a page get read correctly when it matches what's visible on the page. It doesn't replace accurate content, and adding it as a ritual does nothing.

Preballin's rule

Most of this is maintenance. The writing is the smaller part. It's a short list of checks that has to happen every week, and that everybody skips when the week gets busy. That's exactly the kind of repeated work a small system can prep for review without letting anything publish itself. It watches your pages, your unanswered customer questions, and your business details for drift, then hands you a short list of what changed and what to fix. A person still approves every change.

For the cost and scope side, I wrote about how much AI automation costs and how to turn repeated tasks into safe automation rules. You can also see the systems I build or browse the portfolio.

FAQ

Does AI search actually send customers to local service businesses?

It can send visits, and here it did, at a small scale that grew over several weeks. Whether those visits turn into customers is a separate question analytics can't answer. You find out by asking the people who book.

How do I know if it is working?

Two signals, kept apart. Assistant-attributed sessions in your analytics, and new customers telling you an assistant sent them. Only the second one proves the channel brought in business.

Do I need special markup or a special file for AI search?

No. There's no tag, file, or formatting trick that gets a business recommended. Google's own guidance says its normal search practices apply to AI features. The work is plain, careful site and content work: reachable pages, accurate details, a consistent identity, real proof, and clear answers.

How long does it take?

Nobody knows, and anyone quoting a timeline is guessing. Here the referrals showed up gradually across a two-month window, after the page and identity work, and I can't tie them to one change. What holds up is that clear, accurate pages stay useful to every system that reads them.

Can any of this be automated?

The checking can. Finding unanswered questions, catching pages that drift out of date, spotting broken links, and flagging business details that disagree across profiles is all repetitive work. Publishing, price changes, and anything a customer sees stay behind a human approval. The machine checks. You decide.