Local business AEO: you can win on Google Maps and still be invisible in the AI answer

Questions this guide answers

• How do AI answer engines pick local businesses?
• Why doesn't my business show up in ChatGPT or AI recommendations?
• Is optimizing my Google Business Profile enough for AI search?
• How do AI engines answer 'best [business] near me'?
• Where do AI answer engines get local recommendations?

Direct answer

For local recommendation queries — "best personal-injury lawyer in Austin", "good family hotel near downtown" — answer engines do not read your website or your Google Maps pin and hand back a recommendation. They assemble the answer from the third-party surface: vertical directories, review sites, editorial "best-of" lists, and their own grounding data. That is why a business can rank #1 on Google Maps and still be absent from the AI answer.

Your Google Business Profile still matters, but it is the eligibility layer, not the whole answer. In the Whitespark/BrightLocal 2026 ranking-factors survey it accounts for about 12% of AI-citation weight, down from 32% in the classic local pack, while on-page content, reviews, and citations carry more. And no two engines ground local intent the same way, so "optimize for AI" is really three or four different jobs.

Your website is not irrelevant — it owns your objective facts (hours, services, prices) and it converts the click after an engine names you. But it will not win the "best X" recommendation on its own. Closing the gap is a per-city, per-engine loop: Measure where each engine cites you, Diagnose which third-party sources it trusts that you are missing, Execute the fixes, and Verify the answer actually moved.

The Maps-wins, AI-invisible gap

Ranking on Google Maps and being cited in an AI answer are two separate games, and you can win one while losing the other. Maps ranking is governed by Google's proximity, relevance, and prominence model against your Business Profile. An AI recommendation is assembled from whatever sources the engine grounds on — which usually is not the Maps local pack at all.

The gap is documented. In one widely-cited example, cannabis dispensaries that ranked well on Google were effectively invisible on ChatGPT (Cannabis Industry Journal, 2025). The same teardown found local-business profile accuracy of roughly 68% on ChatGPT and Perplexity versus 100% on Gemini — a reminder that each engine works from a different, sometimes stale, picture of your business.

The practical consequence: a dashboard that only tracks your Maps rank will tell you that you are winning while the AI answer quietly recommends three competitors.

Local AI recommendations are built from third-party sources, not your homepage

When you look at where AI citations for local queries actually come from, the weight shifts away from the profile you control toward the sources others control. The clearest published anchor is the Whitespark/BrightLocal 2026 Local Search Ranking Factors survey (published November 2025), which for the first time added an "AI Search Visibility" factor set alongside the classic local-pack weights:

What the weight shift means

Read the two columns together: Google Business Profile drops from 32% to 12%, while on-page content (24%), reviews (16%), citations (13%), and links (13%) carry the AI answer. The lever for being cited in local AI answers is the content and third-party validation an engine can find about you across the web — not a single profile you fill in once.

This is the opposite of the standard local-marketing advice, which still treats the Google Business Profile as the center of gravity. It is central for the Maps pack. It is a minority of the signal once the question becomes which business an answer engine names.

"Just optimize your Google Business Profile" is necessary, not sufficient

Every local-marketing vendor will tell you to optimize your Google Business Profile, and they are not wrong — it is table stakes. A complete, accurate, well-categorized profile is the eligibility layer. It is decisive for Gemini, which grounds heavily on Google Maps and GBP data (though it tends to use that data without citing it directly), and it keeps your name, address, and hours consistent everywhere else.

But eligibility is not selection. At roughly 12% of AI-citation weight, GBP gets you into consideration; it does not win you the recommendation. If your advice stops at "claim and optimize your GBP", you have done the necessary 12% and skipped the majority that decides who the engine actually names.

Each engine grounds local intent differently — there is no shared local index

There is no single "local index" that all the engines query. Each one assembles local answers from its own grounding sources, which is why your competitor can dominate one engine and be missing from another.

The three trust models

An analysis of 6.8 million citations published by Yext (2025) summarizes the trust models bluntly: Gemini trusts what your brand says about itself (about 52% of its citations were brand-owned domains), ChatGPT trusts web consensus and directories (about 49% third-party directory citations), and Perplexity trusts experts and reviews (vertical directories were its single largest source on subjective queries).

Treat the exact percentages as directional — the study is vendor-published — but the shape is consistent with everything else here: you optimize the seven citation signals that all engines share, then do engine-specific work at the margin. For local, that margin is large: the directory and review surface that ChatGPT and Perplexity lean on is exactly where most local businesses have the thinnest presence.

"Near me" is a city-level query to an LLM, not a radius

Local SEO trained everyone to think in radius and proximity. Language models do not work that way. ChatGPT only added opt-in location sharing on 2026-03-31 (Search Engine Land), and it is off by default — so for most users the model infers location from IP, prior conversation, or the city in the prompt, and it rewrites "near me" into an explicit city ("restaurants near me" becomes "top restaurants San Francisco").

Even with location on, precision is coarse. One test of Google's AI Mode in Chicago returned results averaging 4.9 miles away, versus 0.3 miles for Google Maps (Birdeye). The implication for both measurement and content is concrete: model your prompts and your pages at city and neighborhood granularity, not a radius — the engine is answering a city-level question even when the user typed "near me".

Two verticals, worked: legal and hotels

Law firms. Ask an engine for the best lawyer in a city and it fans out across legal directories — Avvo, Justia, FindLaw, Martindale, Super Lawyers, Best Lawyers — far more than across individual firm sites. Martindale-Avvo, a directory operator, describes an "Authority Stack" of four signals it says AI uses to evaluate lawyers: ratings (profiles under 4.0 out of 5.0 get filtered out), reviews, third-party recognitions, and NAP consistency. Treat the specific mechanism as directional — it comes from an interested party — but the direction is clear: a firm with a beautiful website and a thin directory footprint loses to a firm with the opposite profile.

Hotels. Cloudbeds' study of AI hotel recommendations found 55.3% of citations were OTAs such as Booking and Expedia, with recommended hotels appearing heavily on YouTube, blogs, and Reddit, and 72.4% of recommendations going to brand or major-group properties. A separate London teardown (LuxDirect) found that 65.1% of Google AI Mode answers linked to an OTA rather than the hotel's own site, and that a handful of properties captured the majority of recommendations. The lesson for an independent hotel with an excellent website: scale and a good site are not the same as AI visibility, and the citation is usually won on a third-party page.

Both verticals make the same point from opposite ends. Even when you have a strong owned site, your AI citations are won on the third-party surface specific to your category — and that surface is different for a law firm than for a hotel.

Closing the gap: a per-city, per-engine loop

Because the answer is assembled per engine and per city, the work has to be measured the same way. The method is the SolCrys loop — Measure → Diagnose → Execute → Verify — applied to local:

One funnel, not two

One caveat keeps the strategy honest: this is one funnel, not two. The third-party surface earns the mention; your owned site owns the objective facts — hours, services, prices — and converts the click once the engine names you. Optimizing the earned surface while neglecting your own pages, or the reverse, leaves half the funnel on the table. "Your website does not matter for local AI" is as wrong as "just optimize your GBP".

A note on structured data (the honest version)

Structured data helps, but not for the reasons most "AI SEO" posts claim. There is no AI-only schema and no FAQ-schema quota that buys citations — Google has said as much. What schema does is let an engine state your identity and NAP unambiguously and cross-check it against the directories and profiles that cite you. Three patterns carry most of the value:

The highest-leverage property

Across verticals the single highest-leverage property is sameAs — linking your structured data to your Google Business Profile, Yelp, and the vertical directories above, so an engine can confirm it is the same business across every source it trusts (schema.org). Use the most specific type available, give each physical location its own block, and keep the name, address, and phone identical to what the directories show.

Where to start

Start by finding out where AI engines cite you today. SolCrys's free audit shows which sources answer engines pull from for your brand and where the gaps are — free, no credit card (run the audit). If you operate multiple locations or a regulated vertical like legal or hospitality and want help closing the Maps-vs-AI gap per city, talk to us.

Sources

• Whitespark / BrightLocal — Local Search Ranking Factors and the AI Search Visibility factor set
• Yext — AI visibility in 2025: how Gemini, ChatGPT, and Perplexity cite brands (6.8M citation analysis)
• Search Engine Land — ChatGPT enables location sharing for more precise local responses
• Birdeye — The impact of AI on local search
• Cloudbeds — AI hotel recommendations study
• Cannabis Industry Journal — Dispensaries are winning on Google and disappearing on ChatGPT
• Martindale-Avvo — The Authority Stack (directory-operator source; treat as directional)
• schema.org — LocalBusiness

FAQ

Related guides