A water damage company in Phoenix had more reviews than its biggest competitor. Better photos. Bigger ad budget. Even a nicer website. And when people asked ChatGPT some version of "who should I call if water's coming through my ceiling near Scottsdale," the answer kept being the other guys. The agency running the account audited 24 prompt variations before they found the reason, and it wasn't reviews and it wasn't ad spend. The competitor had city-specific mentions on third-party sites. Their client had a great profile and a ghost town around it.
That's local AEO in one story. Getting your local business cited by AI is a city-by-city, engine-by-engine probability game, and almost nobody plays it deliberately yet. We covered the foundations (Business Profile, entity consistency, schema, review volume) in our local SEO for AI search guide. This post is the layer on top: how the engines actually pick local businesses, and how to win cities systematically.
The Window Is Stupidly Open Right Now
Two numbers tell you everything about the moment we're in.
BrightLocal's 2026 consumer survey found 45% of US consumers used an AI tool to find a local business in the past year. A year before that? Six percent. That's a 7x jump in twelve months, and AI is now the third biggest local discovery channel behind Google and Facebook, already past Yelp.
Meanwhile SOCi's 2026 Local Visibility Index, which tested 350,000 business locations, found ChatGPT surfaces only 1.2% of them in local answers. The same locations show up in Google's local pack 35.9% of the time.
Read those together. Demand went up 7x, and the supply side is a nearly empty shelf. When an engine can only confidently recommend one business in a hundred, being that one business is worth a lot, and the bar for becoming it is still embarrassingly low.
How ChatGPT and Claude Actually Pick Local Businesses
The two engines in the title work completely differently, and the tactics change accordingly.
ChatGPT now shows a map panel with business listings for local queries, and the map runs on Mapbox, not Google Maps. Industry reporting puts 60 to 70% of its local place data on Foursquare's Places API. Sit with that for a second. Most business owners haven't touched their Foursquare listing since roughly 2014, if they ever claimed it. The engine deciding whether you exist is reading a directory you forgot about. Beyond places data, ChatGPT leans on Bing-indexed pages, Yelp, and BBB-style directories to verify a business is real.
Claude keeps no local business index at all. It searches the live web per query and reads pages at the passage level, which means one specific paragraph on your site can earn the citation. Otterly's June 2026 study of 379,000 Claude citations found 64% of them point to brand-owned websites, and here's the wild part: Reddit got zero. Not few. Zero.
Real talk: Claude is the easiest engine for a local business to win, because it cites the exact thing you control most, your own pages. ChatGPT makes you go fix your data plumbing first.
Built by the team that got cited in 48 hours.
Content generation, backlink building, AI visibility tracking, and lead capture. One platform, zero guesswork.
See the platform→Build a Per-City Prompt Panel
This is the piece nobody in the local AEO conversation is doing systematically, and it's the whole game for multi-location and service-area businesses.
A per-city prompt panel is a fixed set of buyer questions crossed with every city you serve. Not "near me" queries. Real prompts, the way people actually type them:
- "best [service] in [city]"
- "who should I call if [problem] near [neighborhood]"
- "[service] in [city] that can come out today"
- "is [your brand] any good for [service] in [city]"
Write the panel before you run anything, then test each prompt across ChatGPT, Claude, Gemini, and Perplexity. And run each one 5 to 10 times, because answers are probabilistic and a single run is a coin flip pretending to be data. What you're measuring is citation rate: out of ten runs, how many name you?
The output looks like this:
| City | ChatGPT | Claude | Who wins instead |
|---|---|---|---|
| Austin | 7/10 | 9/10 | you |
| San Antonio | 2/10 | 4/10 | regional chain |
| Waco | 0/10 | 0/10 | nobody consistent |
That Waco row is the good news, by the way. When no business is cited consistently, the seat is empty. We built our prompt discovery process for exactly this kind of panel work, and funny enough, the thing we see across our customer base is that the city where a brand gets cited most is frequently not its headquarters city. Nobody ever guesses their own map correctly. That's why you measure.
Your Reviews Are Invisible (The Text Problem)
Slight detour, but this matters more than almost anything on your website.
A local SEO practitioner working with pizza restaurants found their best client, thousands of reviews at a steady 4.7 stars, completely absent from AI answers while a newer competitor kept getting named. The difference was the words. The competitor's reviews said things like "wood-fired crust" and "go-to spot for date night in [neighborhood]." The client's reviews said "Great service!" a thousand times.
A star average is machine-readable. It's also a single number with no story in it, and language models quote stories. A 4.7 with no text is a rating to Google and a blank to an LLM.
So change what you ask for. Instead of "please leave us a review," ask customers to mention the specific service and their neighborhood. And reply to the blank ones with descriptive text yourself, since your replies are crawlable too. Forty reviews that read like tiny case studies will outperform a thousand empty stars in every AI answer that matters.
See your first AI citation report in under 5 minutes.
No setup calls. No onboarding meetings. Connect your domain and see where AI mentions your brand right now.
Get started→
Winning Cities Where You Don't Have an Address
Service-area businesses ask me some version of this constantly: how do I get cited in Austin when my office is in Dallas? What the engines actually want is evidence tied to the city, not a mailing address.
Four things move the needle, roughly in order:
- Real city pages. Not the find-and-replace kind where only the city name changes. A page that earns a Claude citation names neighborhoods you've worked in, actual response times, local licensing details, and a project or two from that metro. Passage-level retrieval rewards one genuinely specific paragraph over ten interchangeable ones.
- City-level third-party mentions. One consultant watched a client with zero LLM presence start appearing in ChatGPT, Gemini, and Google AI Mode within five days of landing in a single "best [service] in [city]" roundup plus a few community mentions.
- Fix the forgotten directories. Claim and update Foursquare. Check what Mapbox shows for your business. Boring? Extremely. But that's 60-plus percent of ChatGPT's local data layer.
- Cheap, city-tagged PR. Practitioners in the local AEO trenches use low-cost press release distribution to plant service-plus-city signals. Not glamorous, works anyway.
The encouraging stat here comes from a Yext study: 86% of the sources AI cites for local queries are ones you influence, split between your own site at 44% and listings you can edit at 42%. Only 6% is genuinely out of your hands. I mean, compare that to trying to rank in another city's map pack, where you basically can't.
The Leads You'll Never See in Analytics
One agency kept noticing the same weird pattern: ChatGPT referrals barely registered in their analytics, but on sales calls, lead after lead said "I found you on ChatGPT." One service business closed a $26,000 contract from a client who did exactly that. Nothing in the dashboard. The client asked ChatGPT for a shortlist, got a name, and picked up the phone.
That's the local AEO attribution trap. AI-influenced buyers often never click anything, so your analytics quietly tells you the channel doesn't work while it fills your calendar. Two fixes: ask "how did you hear about us" on every single new lead, no exceptions, and use lead capture with source attribution so the clicks that do happen get tagged to the right engine instead of hiding in "direct."
Run It Like a System, Not a Stunt
Honest time budget, if you're doing this manually: building the prompt panel takes an hour or two, and each full run across four engines and five cities at 5-plus repetitions is 3 to 4 hours. Every two weeks. Add the review workflow, the city pages, the directory cleanup, and the odd press release, and the first month runs 15 to 20 hours with 6 to 8 hours monthly after that. Per brand.
You can grind that out yourself, and if you're a single-location shop, honestly, you should at least run the first panel by hand. It's eye-opening. Or RankControl tracks your local-intent AI citations city by city continuously, across every major engine, and flags when a metro flips. The panel never gets skipped because someone got busy.
Your competitors are building backlinks while you read this.
Organic outreach, social mentions, exchanges, and done-for-you link building. Four ways to grow your domain authority on autopilot.
Start building links→
The map at the top of this game is still mostly dark. ChatGPT can name one local business in a hundred, your competitors are still arguing about star ratings, and every city where no one is consistently cited is a seat nobody claimed yet. Go find out which of your cities are lit before someone else reads a post like this one.
