How AI picks the businesses it names near you.

Local AI searchis the new front door of local discovery: a customer asks an assistant who to hire nearby, and the assistant answers with a short list of named businesses. Those names are assembled from a readable, checkable stack — your business profile, your reviews, your name-address-phone consistency, your structured data, and what third parties say about you. Businesses that keep that stack clean get named; businesses that don't are invisible, no matter how good their work is.

The shift, measured

This stopped being hypothetical in the last two years. The most direct measurement comes from BrightLocal's consumer survey work:

On Google itself, the answer now frequently arrives before the listings do — especially for the exact phrasing local customers use:

The practical meaning for a local business: the map pack and the blue links still matter, but a growing share of your future customers see a composed answer first. In that answer you are either named or absent — there is no rank five. (Why a business ends up absent is its own guide.)

How a “best X near me” answer gets assembled

Different engines assemble local answers from different places, and the differences decide where your effort pays off first:

• Google AI Overviews and AI Modework from Google's own local stack — Business Profiles, reviews, the Knowledge Graph, and regular organic results. If your profile is thin or your categories are wrong, Google's AI surfaces feel it first. The full playbook: Google AI Overviews optimization.
• ChatGPT blends what its model remembers about your business with live web search over review sites, directories, and community threads. Two things are worth getting right here. First, the crawler: OpenAI names OAI-SearchBot as the one behind ChatGPT Search, so a robots.txt that blocks it removes you from that surface — while blocking GPTBot, the training-data crawler, does not. Second, the index: OpenAI does not publish which one sits behind those results, but outside testing found roughly 87% of ChatGPT's citations also rank in Bing's top results (Seer Interactive, 2026). That is a correlation measured from the outside rather than a documented pipeline — and an overlap that large still makes getting indexed in Bing worth the hour it costs.
• Perplexityretrieves and cites sources for every answer — local news, “best of” roundups, community threads. It's the engine where a single strong third-party mention shows up fastest (how Perplexity picks sources).

The local stack: what AI actually reads

Strip the mystique and local AI visibility runs on five things a business controls or influences directly:

1. Google Business Profile.Complete, correctly categorized, actively maintained. It feeds Google's AI surfaces directly and anchors the identity every other engine cross-checks.
2. Reviews — volume, recency, and specifics.Engines quote judgment they can attribute: “patients mention gentle sedation dentistry” comes from real review text. A steady stream of specific reviews beats a wall of five-star one-liners.
3. NAP consistency.The same name, address, and phone everywhere — site, profile, directories. Mismatches read as identity confusion, and an engine that isn't sure who you are names someone it is sure about.
4. LocalBusiness schema — the right subtype. Dentist, Attorney, AccountingService, MedicalSpa-adjacent types — not a bare LocalBusiness. Set expectations correctly on this one: Google states that no special markup is required to appear in AI Overviews or AI Mode, and there is no “AEO schema” that buys a citation. What the right subtype does buy is parsing certainty — your services, service area, hours, and price range stated in a form that leaves a machine nothing to infer — plus the classic rich results markup has always earned in ordinary search.
5. Answer-shaped pages for real buyer questions.A page that directly answers “how much does an implant cost in Mesa” is quotable; a services brochure is not. This is where most local sites have nothing to quote.

The order matters: the first four make you legible; the fifth makes you quotable. Legible-but-not-quotable businesses get mentioned occasionally; quotable ones get cited. The step-by-step version of this work lives in how to get mentioned by AI.

The same stack, different stakes by industry

The mechanics above are universal; what changes by vertical is which questions carry money and which trust signals decide them. We keep a dedicated guide per industry:

• Law firms — practice-area questions with the highest stakes per answer, under bar advertising rules.
• Dentists — routine care versus the implant and Invisalign questions that fill a book.
• Med spas — treatment-level questions where trust and reviews decide everything.
• Accountants & CPAs — seasonal question cycles and trust-heavy selection.

Checking where you stand (the honest way)

You can check your own market in ten minutes: ask ChatGPT, Gemini, and Perplexity three to five buyer questions phrased like a real customer — “best X in your town,” “who should I call for Y near Z” — and write down every business named. Two honesty caveats: answers drift between sessions, and one check is a snapshot, not a trend. (The five-prompt method, with what each outcome means, is in does ChatGPT know your business.)

A disclosure, in passing: AEO Action publishes this guide, and checking at scale is the product's job: daily scans that ask the live engines your buyers' questions, the pages and schema from the stack above prepared for one-click approval, IndexNow submission to Bing for the pages that ship, and a dated receipt when an engine starts citing them. If you would rather compare that against hiring the work out, the AEO services breakdown prices both sides. The mechanics on this page are true whoever runs them — by hand, they simply take a calendar, not a weekend.