How a local business gets recommended by ChatGPT

What has changed

Customers used to type “garage Sion” into Google and scroll a page of results. Today, a growing share of people ask ChatGPT, Perplexity, Gemini or Google's AI answers instead. The difference is stark: AI doesn't show ten links, it recommends a few names. If you're not one of them, you don't exist for that customer.

The good news: AI assistants don't recommend at random. They search the web when the question is asked (ChatGPT and Perplexity run real searches), then summarise what they find. What they find is something you can influence.

How an AI chooses who to recommend

Asked for the “best {trade} in {city}”, an assistant with web search does roughly this:

1. It runs one or more searches (often in English, even for questions in other languages).
2. It reads the top pages: directories, guides, articles, forums, business websites.
3. It keeps the businesses mentioned by several independent sources, with quality signals (reviews, awards, years in business).
4. It checks that each business matches the trade and city requested.
5. It writes a short list, with one line of justification per name.

The upshot: AI recommends businesses that others talk about, not the ones that talk most about themselves. Your website matters, mostly as proof that what others say about you is true.

The 7 actions that matter, in order

1. Measure first

Ask AI assistants your customers' real questions, several times (answers vary from one run to the next). Note whether you're named, in which position, and who is named instead of you. That's what the free check does: without measurement, you can't tell whether your efforts work.

2. A complete, consistent Google Business Profile

Gemini and Google's AI answers rely directly on Google Maps data, and other assistants often read directories that reuse it. Exact name, precise primary category, opening hours, phone, website, photos and above all a steady flow of reviews. Details in our guide Google Business Profile and AI.

3. Be where AI gets its information

In our analyses, AI assistants mostly cite local directories, recognised guides (Michelin and similar for restaurants), tourism office websites, local press articles and Reddit threads. Find the sources for your trade and city, and be listed there with identical details everywhere. See Where AI gets its recommendations.

4. A website AI crawlers can read

ChatGPT's, Perplexity's and Claude's crawlers read your site without running JavaScript. If your text only appears after a script loads, or your robots.txt blocks them, they see an empty page. Check in 20 seconds with Your site as AI sees it.

5. One page per service, with the city

“Electric car repairs in Sion” is far more likely to be picked up than a generic “Our services” page. One clear page per key service, with the city, indicative prices if possible, and a short FAQ built from your customers' questions.

6. Quotable proof

AI assistants justify their picks: “4.8 stars from 200 reviews”, “founded in 1998”, “Tesla specialist”. Give them those sentences: figures, years of experience, certifications, awards, visible as text on your site and your listings.

7. Structured data

A LocalBusiness JSON-LD block states unambiguously: here is the name, address, phone and opening hours. Five minutes with our free generator.

What doesn't work

• Stuffing your site with keywords. AI summarises; repetition is ignored.
• Buying fake reviews. Legal risk, and platforms remove them.
• Writing “we are the best”. AI looks for independent sources, not your opinion of yourself.
• Expecting results within a week. AI repeats what the web says about you: when the web changes, answers follow, usually within a few weeks.

How long does it take?

Technical fixes (readable site, structured data) count as soon as crawlers come back. New mentions (directories, articles, reviews) usually take a few weeks to show up in answers. That's why you should measure regularly: once a week is enough to see the trend.