For fifteen years, ranking for "plumber near me" or "best taco truck downtown" meant one thing: get into Google's local pack. Three listings, a map, a handful of star ratings. Win those three spots and you won the query.
That is changing. Local intent searches increasingly get answered before a map ever renders, inside an AI Overview, inside ChatGPT, inside a voice assistant reading out a single recommendation. The searcher does not see ten blue links and a map pack. They see one synthesized answer with two or three sources credited underneath it.
Google's own AI Overviews now appear on a large and growing share of search results, and they behave differently from the local pack they sit above. The local pack pulls listings based on proximity, review signals, and profile completeness. An AI Overview does something else: it constructs an answer first, then decides which sources earned a citation in that answer.
That distinction matters more than it sounds. A business can rank first organically and still lose the click if the AI Overview cites a competitor, a directory, or a review aggregator instead. Visibility inside the answer, not just the ranking underneath it, is becoming the thing worth optimizing for. That is the core idea behind generative engine optimization, or GEO: making sure your business is the one the model chooses to describe when someone asks.
When someone asks an AI assistant "which dentist near me takes walk-ins" or "best coffee shop open now in Austin," the model is not browsing the web live in most cases. It is drawing on a mix of indexed content, structured data, and, for tools with live retrieval, a fresh pull of nearby business listings.
Local citations tend to lean heavily on the same signals that built the old local pack: your Google Business Profile, your review count and recency, and how consistently your name, address, and phone number appear across the web. Those fundamentals did not disappear. They got folded into a bigger evaluation.
What changed is the second layer. Once a model has a shortlist of candidate businesses, it needs language to describe each one: what you actually do, which neighborhoods you serve, what makes you different from the business two blocks over. That description has to exist somewhere the model can read it cleanly. A vague homepage that only says "quality service since 1998" gives the model nothing to cite. A page that says "we install tankless water heaters in the East Side and Mueller neighborhoods, same-day for emergency calls" gives it a sentence to quote.
We wrote about this pattern in more general terms in our piece on the seven factors that drive AI visibility. Local intent is really a specific, high-stakes application of the same mechanics: structure, specificity, and citable clarity beat vague brand copy every time.
There is also a franchise-specific wrinkle worth naming. A national brand with three hundred locations usually has one strong corporate domain and three hundred thin location pages that all read the same. That pattern worked reasonably well for the old local pack, which mostly cared about the Business Profile tied to each address. It works less well for AI answers, because a model trying to distinguish "the Austin location" from "the Denver location" needs each page to say something distinct about that specific location, not a templated paragraph with the city name swapped in.
It is tempting to read "AI is replacing the local pack" as "throw out everything you know about local SEO." That is wrong. The fundamentals are still load-bearing. They just became inputs to a bigger system instead of the whole system.
Google's own structured data documentation for LocalBusiness markup still describes the baseline: business hours, service categories, address, and review data feed knowledge panels and local results alike. That documentation has not been deprecated. If anything, clean structured data matters more now, because it is one of the more reliable ways to hand a model unambiguous facts about your business.
A few things that remain non-negotiable:
None of this is new advice. It is the same advice local SEO consultants have given for a decade, from Moz to BrightLocal's guidance on ranking in AI search. What's different is who is reading the output.
The layer that did not exist five years ago is the one built specifically for machine consumption rather than human browsing. Search engines used to reward pages that ranked for a keyword. AI systems reward pages that answer a specific question in a form that can be lifted cleanly into a generated response.
For a local business, that means writing down, explicitly and in plain sentences, the things a model needs in order to recommend you correctly:
LocalBusiness schema is the technical vehicle for a lot of this, but the schema alone is not enough. The visible page copy needs to say the same thing the schema says, in sentences a model can extract without inference. Redundancy between structured data and plain-language copy is a feature here, not a duplication problem.
This is also where E-E-A-T signals do double duty. If you have not already, it's worth reading our breakdown of how E-E-A-T connects to GEO: the same experience, expertise, authoritativeness, and trust signals that Google has evaluated for years are exactly what an AI model is trying to approximate when it decides which local business is worth naming.
Not always, and that is the uncomfortable part for a lot of local operators. When someone asks a voice assistant or a chat interface "is there a bakery open near me right now," they may get a single spoken or written answer with no map, no list, no second option shown by default.
That is a narrower funnel than the old local pack, which at least showed three choices. Being the one answer cited, or being absent entirely, is a much higher-variance outcome than being ranked fourth on a map that still gets scrolled.
This does not mean the map pack is going away. Plenty of local search still happens inside Google Maps and traditional search, especially for browsing-type queries where the searcher wants to compare options. But the share of "just tell me the answer" queries is growing, particularly on mobile and inside AI-native interfaces. Recent industry coverage of AI Overviews in local search has tracked this shift, noting that businesses can rank first organically and still lose the click if they are not among the sources an AI Overview decides to cite.
The practical implication: you now have two funnels to manage instead of one. The classic local pack still rewards proximity and reviews. The AI answer layer rewards clear, specific, well-structured descriptions of what you do and where you do it. Winning one does not guarantee winning the other.
Voice adds a third layer on top. When a "near me" query comes through a smart speaker or a voice assistant with no screen involved, there is no fallback list to scroll if the first answer is wrong. The model picks one business and reads it out. For categories where voice queries are common, such as restaurants, pharmacies, and emergency services, that makes the stakes of being the cited answer even higher than on a screen-based chat interface.
If you run marketing for a multi-location business, a franchise, or a single storefront, here is where to start:
That last step is the one most local businesses skip, mostly because doing it by hand across five assistants, for every service and location combination, does not scale past a handful of manual checks.
Ranking reports for the local pack are old, familiar territory: rank trackers, review monitors, citation audits. There is no equivalent muscle memory yet for AI answers, and most local marketing teams are flying blind on whether they are actually being cited when a customer asks an assistant for a recommendation.
Bold GEO exists to close that gap. We run a daily refresh across ChatGPT, Perplexity, Gemini, Claude, and Copilot, tracking whether your business gets named for the local and category queries that matter to you, and how that compares to the competitors also showing up in those answers. For a multi-location business, that means visibility broken out by market instead of one aggregate number that hides where you're actually losing ground.
You do not need a big research budget to start. A single scan costs a dollar, and the plans that run ongoing tracking start at $49 a month with a 7-day free trial. For a local business deciding whether AI visibility is worth the investment, that is a cheap way to find out whether you have a problem before you spend a quarter rewriting service pages.
The local pack built an entire industry around getting into three boxes on a map. The next decade of local search will be about getting into a sentence. The businesses that adapt their data and their language now will be the ones an AI assistant reaches for by default when someone nearby asks who to call.
Bold GEO monitors how your brand is cited across ChatGPT, Perplexity, Gemini, Claude, and Copilot on a daily refresh. 7-day free trial, no credit card.