How does a local service business get recommended by ChatGPT, Gemini and Google AI Mode?

Where do AI assistants get their answer to a “near me” question?

“Who can fix a leaking boiler near me tonight?” is not answered from a model's memory. The assistant estimates where the person is, queries place data and the web, and writes a shortlist. Most of that place data sits in a few databases you can claim at no cost.

So the listing is what gets recommended, and your website is where the assistant confirms what a listing leaves out: the exact service, the price, the streets you cover.

Why does an assistant name a competitor instead of you?

Google ranks local results on relevance, distance and prominence (how well known a business is, reviews and links included), and says a better local ranking cannot be bought. Answers built on map data inherit that logic and need words to match: “emergency” or “open on Sunday” can only match text in a category, a service description or a review. The usual gaps:

• A broad primary category (“Contractor”) where a precise one fits (“Roofing contractor”).
• A service listed on the website but missing from the listing.
• Blank or stale hours, so “open now” questions skip you.
• Few or old reviews that never name the work done.
• A service page that is not indexed; the technical SEO checklist guide covers that.

Distance should not be faked: Google's guidelines reject unstaffed virtual offices and keep a service area within about two hours' drive of the base.

Which business facts must match on every listing?

Assistants compare sources. If your site shows one phone number and two directories another, the model may prefer a rival whose details agree. Keep this audit as a sheet and repeat it after any change.

Claim listings you never created and correct them, or ask the platform to remove duplicates: a forgotten one with an old address keeps feeding the wrong fact to every system that reads it.

How do you earn reviews and mentions an assistant can use?

Google says more reviews and positive ratings can help local ranking. Reviews are also text that assistants summarise: in BrightLocal's 2026 survey of 1,002 US adults, 45% had used an AI tool for local recommendations, up from 6% a year earlier, and 82% read AI-written review summaries. “Replaced the pump the same evening in Chorlton” gives an assistant a service, a speed and a place; “great service!” gives it nothing.

• Ask every customer when the job is done, with a direct link, and do not choose who gets asked.
• Suggest what to mention (service, area, speed), never what to say about quality.
• Reply to each review with facts, including what changed after a complaint.
• Earn true local mentions: a supplier's installer list, a club you sponsor, local news.

Google's Maps policy bans incentives for reviews and asking only happy customers. The US FTC rule of August 2024 bans fake and AI-generated reviews and undisclosed insider reviews, with civil penalties. And do not mark up your own reviews: Google recommends review markup only on sites that review other businesses.

What should a service page say so an assistant can quote it?

Open each service page with the facts a person checks before calling:

• What you do and for whom: “boiler repair for homes, most makes”.
• Where: the districts you cover, written out, not only drawn on a map.
• When: hours, emergency cover, the usual wait for a visit.
• How much: a dated “from” price, and what pushes it up.
• What happens next: a call, a message or a booking form.

Write one page per real service, not one per town: Google's spam policies call city pages that funnel visitors to one page doorway abuse. An area page needs facts true for that area, such as response times or jobs done there.

Mark the page up as LocalBusiness or a subtype such as Plumber, using only what the page shows. Markup wins no recommendation by itself; it removes doubt about which business the page describes.

How do you check each month whether AI recommends you?

Local answers depend on where the question is asked: ChatGPT, for one, may use an approximate location from the IP address unless the user shares a precise one.

1. Write 10–12 local prompts“[service] near me”, “[service] in [district] open on Sunday”, “how much is [service] in [town]”, plus one in each language your customers speak.
2. Pin the locationAsk from inside your area or name the district, in a fresh session. Note date and device.
3. Cover five assistantsChatGPT, Gemini, Google AI Mode, Copilot, and Alice AI for Russian-speaking buyers.
4. Log the whole shortlistWere you named, in what position, beside which rivals, and which source did the assistant show?
5. Fix the source, then waitA wrong fact usually traces back to one listing. Correct it, date the change and repeat next month; compare months, not single answers.

Keep calls and website clicks from your Business Profile beside the log, and ask new customers how they found you.

What does it cost: do it yourself, local SEO or GEO work?

Most of this costs time, not money. Pay when you cannot spare the hours or cannot explain what three monthly tests show.

For one location in a quiet market, AI-assisted is enough: AI drafts, a specialist edits and checks. Choose human-led for clinics, law firms, a name shared with another business or several locations: it adds an expert interview and checks twice as many profiles.

When should a local business not pay for AI visibility yet?

• The listing is unverified, suspended or duplicated: fixing that is free and comes first.
• The website is not indexed: start with a technical SEO audit.
• Prices or service area are not settled: no page can state what the business has not decided.
• You have a handful of reviews: spend a month asking every customer first.

Walk away from sellers who promise a slot in ChatGPT's answer, sell review bundles or rent a virtual address to “rank” in another city. None of these systems sells an organic recommendation.