Two numbers describe the state of local AI search, and together they are the biggest visibility gap in marketing right now. Demand: 45% of consumers now use ChatGPT or other generative AI tools to find local business recommendations, up from 6% a year earlier, according to BrightLocal's 2026 Local Consumer Review Survey. Supply: ChatGPT recommends just 1.2% of local business locations, based on SOCi's 2026 Local Visibility Index across more than 350,000 locations. Nearly half the market is asking, and almost nobody is in the answer.
That gap is brutal for the invisible and extraordinary for the visible. Here is why the old local playbook does not transfer, what actually decides the AI shortlist, and the order to fix things in.
Why your Google rankings do not carry over
The reflex assumption is that winning the map pack means winning AI, and the data says otherwise. SOCi found the same brand set that achieved 35.9% visibility in Google's local 3-pack managed only 1.2% in ChatGPT recommendations, roughly 30 times harder, and fewer than half of the businesses leading Google's local results appear in AI local recommendations at all. BrightLocal's survey shows where the attention went: Google's share of local discovery fell from 83% to 71% in the same year AI jumped to 45%.
The mechanical reason is that there is no shared local index. Gemini can draw on Google's own local data, but ChatGPT builds local answers from Bing-indexed pages, directories, and best-of lists, and Perplexity leans on vertical directories, reviews, and Reddit threads. Each engine is running its own version of local search from different raw material, the same fragmentation we documented across engines generally in where AI gets its answers.
The weighting changed too. In the Whitespark and BrightLocal 2026 Local Search Ranking Factors survey, which added AI visibility factors for the first time, the Google Business Profile that carries about a third of the classic local pack drops to roughly 12% of AI-citation weight, while on-page content (24%), reviews (16%), citations (13%), and links (13%) carry the answer. Your profile is the eligibility layer. The recommendation is decided elsewhere.
What actually decides the AI shortlist
Read the research together and four factors do most of the work.
Being describable. An AI cannot recommend what it cannot summarize. Businesses whose sites state plainly what they do, for whom, where, at what price range, with a dedicated page per service, are extractable. Vague homepages that a human would happily click around are dead ends for a model. Experts in the ranking-factors survey put dedicated service pages among the top AI visibility factors.
Reviews, everywhere that matters. Reviews are the trust layer AI leans on for "best" questions, and consumers meet AI answers with review habits already formed: 97% read reviews for local businesses and 41% now always do. The full mechanics of how review platforms feed AI answers are in our reviews deep dive, but the local-specific point is coverage: ChatGPT's local answers draw on the review and directory ecosystem beyond Google, so a Google-only review strategy leaves the other engines reading someone else.
Third-party presence: lists, directories, local press. The single strongest expert-rated factor for AI local visibility is presence on curated best-of lists. Chamber memberships, local news mentions, industry directories, and community coverage are no longer citation-building busywork; they are the source material of the answer. Community threads count too, for the reasons covered in why AI cites Reddit.
Facts that agree with each other. SOCi found only 68% of business contact information shown by ChatGPT and Perplexity matches the details on the business's Google profile, and that AI engines punish name, address, phone, and hours inconsistencies harder than Google does. Wrong hours in an AI answer is not a ranking problem; it is a customer standing outside a closed shop, and it is the local version of the accuracy failure we covered in when AI gets your brand wrong.
The order to fix things in
For a local business or the agency running its marketing, the sequence matters more than the effort.
First, audit what the engines say today: ask ChatGPT, Gemini, and Perplexity the questions your customers ask ("best X in Y", "is Z any good", "X open near me") and record what comes back, who gets named, and which facts are wrong. Second, fix your facts everywhere at once: site, Google profile, directories, socials, one consistent name, address, phone, hours, and service list. Third, make every service its own page with a plain-language answer at the top. Fourth, build review depth and recency beyond Google, and respond, since 80% of consumers say they are more likely to use a business that responds to all its reviews. Fifth, earn the third-party layer: local lists, local press, community presence. Then re-check monthly, because these answers move.
That audit-and-recheck loop is exactly what Diploria automates: it polls the local-intent questions across 11 AI surfaces, stores every answer and citation, flags wrong facts against your verified details, and shows which competitors are taking the recommendations you are missing. The free check takes a minute and tells a local brand immediately whether it is in the 1.2% or the 98.8%.
The honest caveats
Two things keep this in proportion. AI's local answers are young and sometimes wrong, and users know it: 21% of ChatGPT users switch to Google to verify local information after asking. And the map pack still matters enormously; 71% of discovery is not a number anyone abandons. This is not a migration from one channel to another. It is a second shortlist forming next to the first one, with almost no competition on it yet. That last part is the opportunity, and it will not stay true for long.