Being Seen vs. Being Selected: How AI Decides Which Local Businesses to Recommend

The map moved, and proximity stopped winning

For two decades, local search ran on one dominant variable: distance. You ranked in the map pack partly on relevance and reviews, but proximity to the searcher was the gravitational force. A plumber three blocks away beat a better plumber across town. The entire discipline of local SEO was built to optimize the signals around that core, accurate listings, category selection, review volume, and a tightly managed Google Business Profile.

That model is now breaking in the surface most consumers are migrating to. According to BrightLocal's Local Consumer Review Survey 2026, a representative panel of 1,002 US adults, the share of consumers using AI tools to find a local business climbed from 6% in 2025 to 45% in 2026. AI assistants have surged into third place for local discovery, behind only Google and Facebook, and Google's own share of local discovery dipped from 83% to 71% in the same year. The audience is moving, and it is moving fast.

When that audience asks an AI for a local recommendation, distance largely stops mattering. Local Falcon's whitepaper on Google AI Overviews, built on 60,000 real-world simulations across 4,423 businesses in 20 countries, found effectively no correlation between distance and ranking position inside AI Overviews. The correlation coefficient was 0.001. In a traditional local pack, distance is one of the strongest predictors of who shows up first. In AI answers, it is statistical noise.

In Google's local pack, proximity is the dominant ranking force. In AI Overviews the correlation between distance and rank is 0.001. The map collapsed into a recommendation, and the recommendation does not care how close you are.

Proximity has not vanished entirely. Local Falcon found a faint inclusion effect: businesses within roughly one mile appeared in AI Overviews 72.0% of the time versus 68.5% for those slightly farther out. Distance can still get you into the candidate pool. But once you are in, it does nothing to determine whether you are the business the AI names. Content authority and reputation take over completely. This is the single most important shift in local for owners to internalize: the nearest business no longer wins by default, and the better-documented one frequently does.

Visibility is not selection, and the gap is enormous

Here is the structural problem. In classic local SEO, visibility and selection were nearly the same event. If you ranked in the 3-pack, you were seen, and being seen was most of the battle. AI severs the two. Being indexed, listed, and even ranked on Google no longer means an AI will name you.

The most comprehensive measurement of this gap is SOCi's 2026 Local Visibility Index, which analyzed nearly 350,000 locations across 2,751 multi-location brands. The headline numbers are stark. AI platforms recommended only 1.2% of locations on ChatGPT, 11% on Gemini, and 7.4% on Perplexity, compared with 35.9% appearing in Google's local 3-pack. SOCi's own summary: AI visibility is three to 30 times harder to achieve than ranking well in traditional local search.

The deeper finding is the overlap failure. Fewer than half of the brands leading in traditional local search also appeared among the most visible brands in AI results. In retail, only 45% of the top 20 brands by traditional local visibility overlapped with the top 20 in AI. Strong Google rankings did not carry over. Brands that had spent years winning the map pack discovered they were nearly invisible in the answer.

• 1.2% of locations recommended by ChatGPT vs. 35.9% in Google's 3-pack (SOCi).
• AI visibility is 3x to 30x harder to achieve than traditional local ranking (SOCi).
• Only 45% of top-20 traditional retail brands also ranked top-20 in AI (SOCi).
• Local Falcon found Google AI Overviews now appear in 40.2% of local business searches.

Being seen by Google is not being selected by the model. Fewer than half the brands winning traditional local search win in AI. The 3-pack you fought for does not transfer to the answer.

This is the visibility-versus-selection gap that defines the vertical, and it is wider in local than in any other category we have measured. A roofer can dominate the map pack in his city and never be mentioned when a homeowner asks ChatGPT who to call. The work that produced the first outcome does not produce the second.

The signals that get a local business recommended

If proximity is dead and Google rank does not transfer, what does the AI actually use? Four signal classes do the heavy lifting, and they reorder the priorities most local owners are used to.

1. The Google Business Profile, as a data spine. The profile still matters, but its job changed. It is no longer a ranking lever you tune for the 3-pack. It is the structured source of truth that AI systems lean on to confirm a business exists, what it does, where it is, and how it is rated. Gemini achieves 100% business profile accuracy because it is grounded directly in Google Maps data, which is exactly why Gemini recommends local businesses at nearly ten times the rate of ChatGPT. A complete, accurate, fully populated profile is now table stakes for being eligible at all.

2. Review volume, recency, and sentiment, as a filter not a gradient. This is the most misunderstood signal in local GEO. AI does not rank you by star count the way a human might scan a list. It uses reviews as a confidence threshold. SOCi found that locations recommended by ChatGPT averaged 4.3 stars, Gemini 3.9, and Perplexity 4.1. Businesses with ratings near 3.4 stars and review response rates below 5% were effectively invisible, excluded entirely rather than ranked lower. In traditional search, a 3.4-star business could still surface on proximity. In AI, it does not surface at all. Recency and response rate are part of the filter: BrightLocal found 97% of consumers still rely on reviews, 41% now "always" read them (up from 29%), and slow or generic review responses are increasingly read as a red flag.

3. Directory and citation consistency, as a trust signal. The model cross-references your information across the web, and contradictions erode confidence. Business profile data is only 68% accurate on ChatGPT and Perplexity versus 100% on Gemini (SOCi), and that accuracy gap directly explains the recommendation gap between platforms. Consistency is measurable and it pays: Yext's analysis of 6.9 million citations found meaningful correlation between unstructured citation volume and AI search visibility, and Yext data from 620,000 locations showed those syncing 50 to 75% of a 200-plus publisher network saw a 95% average increase in website clicks from Google. NAP consistency is no longer hygiene. It is the difference between a model that trusts your data and one that hedges.

4. Third-party "best of [city]" lists and editorial. When an AI answers "best tacos in Austin," it defers to the publishers humans already trust for that judgment. A placement in a credible local roundup, a city magazine's annual list, or a respected industry directory is worth more than any amount of on-site copy claiming you are the best. The model treats independent editorial as verification it cannot get from your own domain.

AI does not rank reviews. It uses them as a gate. Below the sentiment threshold you are not ranked lower, you are excluded. A 3.4-star business that survives on proximity in the map pack simply disappears in the answer.

Notice the reordering. Two of the four signals, citations and editorial, live almost entirely off your own property. The local owner's job has shifted from tuning a profile he controls to shaping a consensus he can influence but not author.

Why local GEO is not local SEO

The temptation is to treat this as local SEO with extra steps. It is not. The two disciplines share inputs but optimize for different mechanics, and confusing them is why so many well-ranked businesses are invisible in AI.

Classic local SEO accumulates signals to win a proximity-weighted ranking. You pile up reviews, citations, and category relevance, and Google sorts the nearby candidates. Volume and distance do real work. Local GEO is a qualification problem, not an accumulation problem. The AI is not sorting a list of nearby options by score. It is deciding whether your business is a legitimate, well-documented answer to a specific question, then naming a small set of winners.

Three concrete differences follow:

• Proximity is a ranking factor in SEO and a near-irrelevance in GEO. Distance correlation with AI Overview rank is 0.001 (Local Falcon). The map pack rewards the closest acceptable option. The answer rewards the best-documented one.
• Reviews are a gradient in SEO and a filter in GEO. More and better reviews lift you in the pack. In AI, below-threshold sentiment removes you entirely (SOCi). The first dollar of review work, getting above the bar, is worth more than the hundredth.
• Query patterns invert. Local Falcon found that queries without a location name are 11.1 percentage points more likely to trigger an AI Overview than queries with one (46.1% vs. 35.0%), and longer, informational queries trigger AI far more often (queries of 100-plus characters at 60.5% vs. 19.7% for short queries). The detailed, conversational, "near me who can fix X today" question is exactly where AI intervenes, and exactly the kind of query classic local SEO never optimized for.

That last point cuts both ways and creates a defensive opportunity. Because location-named queries trigger AI Overviews less often, businesses that genuinely dominate "[service] [city name]" searches retain more traditional traffic. The exposure to AI is highest in the service categories where consumers research before buying. Local Falcon found the highest AI Overview trigger rates in window cleaning (65.0%), carpet cleaning (64.6%), personal injury attorneys (62.1%), and wedding photographers (61.8%). These are considered purchases with real research behind them, which is precisely where the AI inserts itself between the consumer and the business.

Local SEO accumulates signals to win a proximity-sorted list. Local GEO qualifies you as a legitimate answer, then names a winner. One rewards being close and abundant. The other rewards being documented and trusted.

The structural advantage local businesses are sleeping on

There is good news buried in the disruption, and most owners are missing it. Local businesses may hold a structural advantage over national chains in GEO, for reasons that have nothing to do with budget.

A neighborhood business is already hyperlocal. It understands its market, knows the specific questions its customers ask, and can move in days rather than the months a national brand needs to push a change through committee. More importantly, it has raw material a chain lacks: genuine specialty expertise, named local projects, real customer stories, and the kind of specific, verifiable detail the model prefers over generic national copy. A single coffee shop that explains its roast, sourcing, and brewing method gives the AI more to quote than a chain with a thousand interchangeable listings and generic reviews.

The multi-location finding sharpens this. SOCi's data showed that strong performers were not the biggest brands but the most disciplined ones. Culver's, a restaurant chain, beat category benchmarks with 30.0% ChatGPT and 45.8% Gemini recommendation rates, driven by strong ratings and complete profiles. Liberty Tax, after improving profile coverage, ratings, and data accuracy, reached 68.3% Google 3-pack visibility alongside 19.2% Gemini and 26.9% Perplexity recommendation rates. Meanwhile, brands with strong traditional rankings but weaker data, including some household names, underperformed badly. Execution beat size.

A national brand has scale. A local business has specificity, speed, and genuine expertise. In a system that rewards documented substance over generic abundance, the small operator is not the underdog. The undocumented one is.

For multi-location operators the risk is the inverse of the opportunity. Each location needs consistent, accurate data across every platform the AI touches, and a single location with bad data or a sub-threshold rating can drag the brand's overall trustworthiness. Consistency at scale is the hard part, and it is where most multi-location brands lose.

The execution playbook

The work divides cleanly by business type, but the foundation is shared. None of it is optional once 45% of your customers are asking an AI before they ask you.

Shared foundation: become the most trustworthy data on the web.

• Complete and verify the Google Business Profile fully. Every field, accurate hours, correct categories, real photos, populated attributes. This is the spine Gemini and AI Overviews ground themselves in, and incompleteness is why ChatGPT and Perplexity get you wrong 32% of the time.
• Fix citation consistency across the directory ecosystem. Identical name, address, and phone across Google, Bing Places, Apple Maps, Yelp, and the major directories. Contradictions make the model hedge, and hedging means exclusion. Yext's 6.9M-citation data ties citation volume and consistency directly to AI visibility.
• Get above the review threshold, then stay recent and responsive. The first job is clearing the sentiment gate, roughly the 4-star band where AI starts recommending. Solicit reviews continuously so recency stays fresh, and respond to all of them. Response rates below 5% correlate with invisibility (SOCi); slow responses are now a consumer red flag (BrightLocal).
• Add LocalBusiness schema, matched to your subtype. Restaurant, AutoRepair, Attorney, and the rest, with name, address, phone, and hours that match every citation exactly. When schema contradicts your other data, Google discounts it entirely, so consistency beats cleverness.

Single-location businesses: out-specify, don't out-spend.

• Document specific local expertise the model can quote: named projects, regional content ("winterizing pipes in Minnesota"), transparent pricing for common jobs, before-and-after results, and staff credentials.
• Win the third-party "best of [city]" placements and local press that AI defers to. One credible editorial mention outweighs a page of self-description.
• Write for the long, conversational, informational queries that trigger AI most, in plain question-and-answer form.

Multi-location brands: enforce consistency, then differentiate locally.

• Centralize location-data management so every profile is complete and identical in its core facts across every platform. One bad location erodes brand-level trust.
• Lift the laggards. SOCi's evidence is that improving the weakest locations' ratings and data accuracy moves the brand's whole AI visibility, as Liberty Tax demonstrated.
• Give each location genuine local substance, not boilerplate. Localized content beats templated pages the model reads as thin and duplicative.

Everyone: measure selection, not just rankings.

Classic rank tracking tells you where you sit in the map pack. It cannot tell you whether ChatGPT, Gemini, or Perplexity named you when a customer asked, which is the event that now precedes the visit. You need to track share of voice in AI answers, the local prompts you appear and do not appear in, and which competitors get recommended when you do not.

The old job was ranking near the searcher. The new job is being the answer the engine trusts enough to name. Proximity was a placement you earned with distance and volume. Selection is a reputation you earn with accuracy, sentiment, and proof.

The local businesses winning this transition are not the closest or the biggest. They are the ones whose data is the most consistent, whose reviews clear the gate and stay fresh, and whose expertise is documented specifically enough for a machine to repeat. The map moved. The question is whether the engine trusts you enough to recommend you when the customer never sees the map at all.

Every statistic in this piece follows Visigent's research standards: named sources, dated figures, corrections on request to help@visigent.com.