Voice Search for Local Businesses: How Assistants Pick One Name

It is 7:40 on a Saturday morning in Alamo Heights. A man bites down on a popcorn kernel left over from movie night and feels a molar give. He does not open a browser. He picks up his phone and says, "Hey Google, find a dentist open today near me." The phone thinks for about a second and answers with one office, its distance, and an offer to call. He says yes. The receptionist who picks up books him for 10:15.

Four other dental offices within three miles were open that morning. None of them was mentioned. None of them will ever know the patient existed, because there is no impression report for a name that was never spoken.

That is the basic arithmetic of voice search for a local business. A typed search shows a map with three businesses and a list below it. A spoken answer usually names one, sometimes reads two or three if the person asks for options, and then waits for a decision. The decision is often "call them," and the assistant places the call itself. The customer never sees your website, your photos or your clever tagline. The assistant reads a few fields from a listing, and the listing either earns the call or hands it to someone else.

That changes what "being found" means. On a screen, a business can survive in fourth place on the strength of a good photo and a nearby pin. Out loud, ranking stops being a ladder and becomes a coin with one side.

No assistant publishes its formula for local answers, and anyone selling you the exact recipe is selling a guess with a logo on it. What is public turns out to be useful enough.

Google states plainly that its local results weigh three things: relevance (does this business match what was asked), distance (how far it is from the person or the place they named) and prominence (how well known and well reviewed it is). Google's assistants draw on the same Maps data, so it is reasonable to assume the same three forces are at work, with one more filter on top: the assistant has to be confident enough to say a single name out loud.

Apple says much less about how Siri and Apple Maps rank places. Amazon says less still about Alexa. But when you listen to the assistants answer the same questions over months, the pattern is steady enough to plan around.

Relevance comes from categories and details

An assistant asked for an "emergency plumber" looks for businesses whose category is plumber and whose listing signals urgent or after-hours service. A listing that says only "contractor" is not in the running, however good the plumber. Categories, services and attributes are how a machine learns what you do.

Distance comes from your pin and the customer's phone

The assistant knows where the question was asked. You cannot move your customers, but you can make sure your map pin sits on your actual door and your service area is stated wherever the platform allows it.

Prominence comes from reviews and the rest of the web

Ratings, the number and freshness of reviews, and how often your business is mentioned elsewhere all feed the sense that you are an established choice. Assistants often say "highly rated" out loud, which is a polite way of telling you what they sorted by.

Confidence comes from agreement

This is the part most advice skips. Before an assistant names you, it needs to trust the facts it is about to read: that you exist at that address, that you are open right now, that the number connects. When Google, Apple, Yelp and Bing all describe the same business the same way, that trust is cheap. When one says you close at 5 and another says 7, the assistant has a reason to pick a competitor whose story is simpler.

Here is the useful surprise. The assistants are many, but the sources are few. Most local voice answers trace back to one of four listings, and an owner can claim all four in an afternoon.

Each source gets its own treatment in this series. The Apple side, which is the most neglected, is covered in how to get your business on Siri. The Google side, with the most fields and the most ways to go wrong, is in how Google Assistant and Gemini answer local searches. Alexa's more tangled supply chain gets its own explanation. If you are only going to read one thing about the Google map itself, make it how to show up in Google Maps.

Your website sits behind all four. Assistants use it to confirm what a listing claims, and to answer the questions a listing cannot: what a water heater swap costs, whether you take a particular insurance plan, how long a crown appointment runs. A site that contradicts your listings undercuts them. A site that agrees with them, in plain text a machine can read, backs them up.

One quirk worth knowing: owners tend to keep the listing they personally check, usually Google, in good shape and let the others drift. We wrote a short piece on the Google setting that decides whether Siri recommends you, which is really about that blind spot.

Picture a taco shop on the West Side that moved two blocks a few years ago, switched phone carriers last spring, and started closing at 2 p.m. on Mondays. The owner updated Google the day of each change, because Google is the one she checks. Apple Maps still shows the old address. Yelp still has the old number. Bing never heard about Mondays.

Each of those listings is "mostly right." Together they describe three different restaurants. An iPhone user asking Siri for breakfast tacos nearby may be routed to an empty storefront. An Alexa user may hear a number that rings in somebody's kitchen. Nobody complains. They just eat somewhere else, and the owner wonders why Mondays are slow.

The repair is dull and permanent: one canonical record, written down once, copied everywhere, and updated everywhere on the same day.

Not every voice request is "call someone." Plenty begin as questions. How much does it cost to fix a slab leak? Do I need a referral to see a dermatologist? Can you bring a dog onto the patio at the brewery on Broadway? For these, the assistant may read an answer from a website, sometimes naming the source and sometimes not.

The pages that get read aloud share a recognizable shape:

1. A heading that is the question, worded the way a person would say it, with the city in it when the answer depends on place.
2. A first sentence that answers, in roughly thirty words, with the number or the yes or no up front.
3. Detail underneath for the reader who keeps going: what changes the price, the exceptions, the next step.

Here is the difference on a San Antonio plumber's site. The weak version opens with "Our Services" and buries the price in paragraph four, after a history of the company. The strong version has a heading that reads "How much does slab leak repair cost in San Antonio?" and, right under it, a single sentence with the plumber's real price range, the one thing that moves it, and how long the job usually takes. Everything else follows.

Structured data helps machines find those pairs. FAQPage markup labels which text is a question and which is its answer. LocalBusiness markup restates your name, address, phone and hours in a form software reads without guessing. There is also a "speakable" property that marks a passage as suitable for reading aloud, but Google supports it only in a limited beta for news publishers in US English, so for a dental office or a roofing company it is a curiosity, not a lever. The voice search schema guide covers what to mark up and what to skip, and near me voice searches covers how people actually phrase the request.

You cannot manage what you never hear. The assistants send no reports, so you have to ask them yourself, the same way, on a schedule. This routine takes about an hour a month and catches most problems before they cost you a season.

If you would rather start from a baseline than a blank spreadsheet, the free Voice Search Visibility Check looks at your business across Google, Apple Maps, Yelp and your website, and flags the details assistants read. For the Google map side in depth, the free Map Pack Audit goes field by field through the profile.

The sources are the same for everyone. What customers ask, and where they are standing when they ask it, is not.

Restaurants live on hours, menus and "open now," and a large part of their voice traffic comes from people who are already hungry and already driving. Voice search for restaurants covers holiday hours, order links and the problem of dish names an assistant cannot pronounce.

Home service companies (plumbers, HVAC, electricians, roofers) get the most urgent requests and the most "call them" commands, so the phone line and the after-hours details matter more than anything on the website. See voice search for home services.

Medical and legal practices field questions that carry more weight: insurance, referrals, specialties, whether new patients or clients are being accepted. Accuracy there is a professional duty as much as a marketing one. Voice search for medical and legal practices covers what to publish and what to leave for a human on the phone.

The device changes the moment. A driver using CarPlay or Android Auto wants a call or a route, right now, without looking at anything, and voice search in the car walks through that situation. Smart glasses, earbuds and the newer always-listening assistants push even further from the screen, which is the subject of ambient assistants.

If you are starting from nothing, read the two big platforms first, then Alexa, then whichever guides match how your customers ask. If something is already broken, skip straight to the fixes.

If you would rather hand the whole thing off, voice and ambient assistant optimization is the done-for-you version: every listing claimed and matched, the key pages rewritten for the ear, and the monthly test run and logged. It sits beside our local marketing work, because it is the same record kept accurate. For the typed side of the same shift, where assistants write paragraphs instead of saying one name, start with the AI search guide and its short primer on optimizing a website for voice search.