Retrieval, then selection — in the local case
Local queries have a distinctive shape: they bind a category to a place. "Best HVAC company in Boise," "real estate auctioneer near Charlotte," "who does estate sales in Des Moines." An assistant answering one of these first retrieves candidate businesses from sources it can pull — maps entities, directory listings, review corpora, and web pages that name a business alongside its category and city. Then it selects a short list to actually recommend.
You lose at the retrieval step, not the selection step. If your name, address, and phone disagree across the web, or your category is ambiguous, or nothing citable ties you to the city and the service, you never enter the candidate set — so selection never gets a chance to name you. The fix is upstream: make the local entity clean and retrievable before worrying about being chosen.
Which signal each local query shape leans on
| Query shape | Signal it leans on |
|---|---|
| "best [category] in [city]" | Clear category + city binding in citable pages; review presence |
| "who should I hire for [job] near me" | Consistent NAP and service-area clarity so the entity resolves |
| "[category] recommendations [neighborhood]" | Local content that names the area and answers the intent |
| "is [business] any good" | Visible, parseable reviews and a stable entity identity |
| "top [category] companies [region]" | Being named alongside peers on pages a retriever trusts |
Signals are qualitative levers, not a ranked scorecard. The point is which input each shape depends on to retrieve you.
Measuring it means running the real queries
There is no single visibility score to look up. Answers shift between runs and differ across engines, so the honest instrument is to run the actual queries and count outcomes. The AI Citation Institute uses a frozen five-engine panel — ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews — and samples each city-and-category query k times per engine. That produces a named rate: how often you get recommended, per engine, versus the competitors your customers are choosing between.
One lever we can quote with a measured number is internal linking. In a 91-post audit on 2026-07-11, pages with zero inbound internal links were self-cited about 4.5% of the time; pages with ten or more inbound links reached roughly 44%. Local content that is well cross-linked and clearly tied to your city and category is far more retrievable than an orphaned page.
We do not publish a specific local-business client delta — we don't have one filed. The mechanic and the instrument are what we stand behind.
Common questions
Is AI visibility just local SEO with a new name?+
No. Local SEO optimizes for ranked links on a results page. AI visibility is about whether an assistant retrieves your business as a candidate and then names it in a written answer. The signals overlap — consistent NAP, category clarity, reviews — but the outcome is being cited in prose, not placed in a list.
How do you measure whether AI recommends my local business?+
By running the actual city-and-category queries your customers ask across a frozen five-engine panel — ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews — and sampling each query several times per engine. You count how often you get named versus competitors. A single run is noise; repeated sampling is the measurement.
What is the fastest thing a local business can fix?+
Reconcile your name, address, and phone so they match everywhere, and make your category and service area unambiguous. If an assistant cannot resolve which business you are or what you do where, it cannot retrieve you as a candidate — so nothing downstream matters until that entity is clean.
keep reading
See what AI says about your brand right now — free, on a live engine.
Scan your domain