The way people search is changing

For a local service business, getting found has always been the whole game. For twenty‑five years the playbook was clear: rank on Google, complete your Google Business Profile, pile up reviews, hold your spot on the map. Nearly half of all Google searches carry local intent, and roughly three out of four people who search “near me” visit a business within a day — so the stakes were never small. Show up, and the phone rings.

That playbook still matters. But it is no longer the only front door. A fast‑growing share of people now open an AI assistant and ask a full question in plain language — “who’s a reliable plumber near me that can install a tankless water heater?” — and get back a short answer with a name or two. Recent industry surveys already put it near 45% of consumers using AI tools to find local businesses. They don’t scroll ten results. They act on the name the model gives them.

And they ask differently. Google’s own 2026 data shows AI‑era queries running about three times longer than the old two‑ or three‑word search — more specific, more conversational, more likely to name a neighborhood, a specialty, or an exact problem. For a service business that’s the gap between “electrician” and “licensed electrician for a panel upgrade in East Nashville.” The more specific the question, the more the details of who you are decide whether you’re the one named.

How it works: one local search is sent to ChatGPT, Claude, and Perplexity; each returns which businesses it named and the public signals behind each match; the results become a dated AI Search Snapshot.

one search  →  three AI models  →  one dated snapshot

AI doesn’t answer the way Google did

You already know the local‑search world. You’ve spent years getting the Google Business Profile complete, the reviews flowing, the map pin ranking — maybe running Local Services Ads with the Google Guarantee badge. All of that handed you a ranked list to compete on, and let the customer choose from it.

An AI model works differently. It hands back an answer and, often, just a name or two — which means the model chooses. You’re no longer fighting for position three on the map. You’re trying to be the business the model actually says out loud.

“Visible to AI” is not the same as “ranks well on Google.”

A service business can dominate the local map and never once get named by an AI model — and a smaller shop Google barely surfaces can get mentioned across every model. Different systems, different rules.

And “visible” is meant literally. In some cases an AI system may not access, read, or use a website the way a traditional search crawler does — because of how the site is built, how it’s configured, or simply what the model can reach. Some sites are, in effect, invisible to the models customers now ask.

Browse search snapshots

So what does “matching” even mean?

When someone asks an AI a very specific question — a service business in one particular part of town, offering one specific service, with a certain kind of reputation, licensed or bonded or guaranteed in some particular way — the model returns whatever it decides best matches. The more specific the question, the more those nuances start to matter.

But matches on what, exactly? The words on your own website? Your Google Business Profile or a directory listing? Customer reviews? A “best of” article someone else wrote? Project photos? No model hands you its exact formula — but the signals it responds to are becoming clearer, and mapping them is exactly what the whole local‑search industry is racing to do right now.

What sits behind a single match

Every snapshot shows the public signals the AI pointed to

Where the model says it found the business, the type of page, the sources it cited, and the entities it picked up — all recorded, none invented.

Anatomy of a match: AI match confidence, where the AI says it found the business (website, local directory, business profile), the sources it mentions (article, review platform, portfolio), the page type, other cited signals, key entities detected, and schema data.

So we started writing it down

This site is that record — the searches we run, published as dated snapshots.

Every day we take real local searches — the kind people actually put to an AI — and ask the leading models the exact same question. We note which businesses each one names, which names come up more than once, and the public pages and signals each model points to as support. We confirm the businesses are real. Then we publish it as a dated snapshot anyone can open.

We also classify the kind of question being asked, because the phrasing shapes the answer. Some are broad — “plumbers in Cleveland.” Some are highly specific — “electricians who install solar panels in Albuquerque.” Some pin the location down to city and state — “roofing contractors in Lexington, KY” — while others name a place more loosely, a neighborhood or a metro — “coffee shops in Wicker Park” or “movers in the Phoenix metro.”

One search, captured across three models on a single day, is a snapshot. Thousands of them, gathered over time, become a living record of how AI answers real questions about local business.

1

We ask

The same local search goes to OpenAI, Claude, and Perplexity — word for word.

2

We record

We log which businesses each model named and the public source it said supported the match, and we check the businesses are real.

3

We publish

Every result becomes a dated snapshot you can read, compare, and link to.

A note on method. Every model gets the same neutral prompt — asked to name the businesses first and identify a source second — kept identical across all three models and every day it runs. So the wrapper sits as lightly as possible, and any difference you see is the models moving, not us. View the full methodology →
Browse search snapshots

The search snapshots

This is the heart of it. Every day we put real local‑services searches — the kind homeowners and property managers actually ask — to ChatGPT, Claude, and Perplexity, then record which businesses each model names and the public signals behind them. Explore every snapshot below: search a query or a place, or filter by profession, the type of question, and how the location was framed.

Loading the snapshots…

What should we test next? Enter a local search idea — example ideas include piano lessons for kids in Madison, estate planning lawyer in Boise, bird vet in Portland, and best vegan bakery in Richmond.

Want us to test your market?

Suggest a local search you want us to run, or get a free read on whether AI can see your business today.

Request a free AI Visibility audit for your business