A quiet shift is happening in how customers find local businesses. Instead of scrolling a page of results, a growing share now asks ChatGPT, Perplexity, or Google’s AI a question — “who’s the best dentist in Franklin?” — and gets a short, confident answer naming two or three businesses. No page two. No “maybe.” For that customer, the decision is mostly made before any website gets visited.
The numbers behind this are stark: roughly 36% of local businesses show up somewhere in Google’s local pack results, but only around 1% ever get named by ChatGPT. That gap is the opportunity. The AI answer slots for your market are, most likely, sitting unclaimed — and the businesses that claim them first tend to keep them, because AI engines keep citing the sources they already trust.
So how does an AI decide who to recommend? Not by magic, and not by ad spend. Six signals, all documented, all fixable.
1. Consistent citations — the same name, address, phone, everywhere
AI engines cross-reference. Before naming a business, they effectively check whether the story holds up across Google, Yelp, Facebook, industry directories, and the open web. If your suite number is different on Yelp, your old phone number lingers on a directory you forgot exists, and your business name has three variants floating around, you read as uncertain — and engines don’t recommend uncertain.
This is the least glamorous fix on the list and one of the most important: audit every listing, correct the inconsistencies, fill in the directories you’re missing from.
2. Bing — yes, really
ChatGPT leans heavily on Bing’s index for anything current or local. Almost every local business optimized for Google years ago and hasn’t thought about Bing since — which means Bing Places, Bing’s equivalent of a Google Business Profile, is usually empty or auto-generated junk for your market. Claiming and fully building out your Bing Places profile is the single cheapest AI-visibility win available right now, precisely because nobody does it.
3. Reviews — quantity, rating, recency, and responses
Businesses that AI engines recommend average around 4.3 stars. Below roughly 3.5 — or with reviews that stopped a year ago, or with no owner responses — you’re effectively invisible in AI answers. Engines treat an active, well-reviewed, well-responded profile as evidence a real, functioning, trusted business exists.
You can’t buy reviews (and shouldn’t try — fake-review patterns are exactly what these systems are good at spotting). What you can do: make asking systematic. A QR card at checkout, a follow-up text after every appointment, and a response to every review — including the bad ones, especially the bad ones.
4. Schema markup and crawler access — the label on the box
Schema markup is a block of structured data on your website that tells machines, unambiguously: this is a business, here’s its name, category, address, hours, services, and prices. Humans never see it; AI crawlers rely on it. Most local business websites have none.
Related and even more basic: some websites accidentally block AI crawlers in their robots.txt file, usually because of a template default. If GPTBot can’t read your site, ChatGPT is choosing from your competitors.
5. Content that answers questions the way people ask them
AI engines quote sources that sound like answers. A services page that says “Welcome to our state-of-the-art facility” gives an engine nothing to work with. A page that plainly answers “How much does Botox cost in Scottsdale?” or “How long does lip filler last?” — with specific, authoritative, current information — is quotable. Quotable gets cited.
6. Third-party mentions — the sources AI already trusts
A large share of what AI engines cite for “best X in Y” questions isn’t the businesses’ own websites — it’s local press, “best of the city” roundups, and industry listicles. Find the lists AI already cites for your market (just ask it, and check the sources it links), then work on getting into them. Many accept submissions; some just need an email.
What you can do this week, free
- Run your baseline. Ask ChatGPT and Perplexity the five questions your customers would ask. Screenshot the answers. This is your “before” — and your proof later that the work worked.
- Claim Bing Places. Thirty minutes, and you’re ahead of nearly every competitor.
- Check your robots.txt (yoursite.com/robots.txt) for anything blocking GPTBot, ClaudeBot, or PerplexityBot.
- Respond to your last ten reviews. Recency of engagement counts, and it’s free.
One honest caveat: none of this moves overnight. AI engines re-crawl and re-weigh on their own schedule — expect 30 to 60 days before answers shift, which is exactly why the right move is to fix the signals once, properly, and then measure.
Want all six signals fixed for you — with proof?
The AI Visibility Package does everything in this post, done-for-you: baseline AI test with screenshots, directory and citation cleanup, Bing Places, schema markup kit, Business Profile tune-up with a review-generation kit, answer-ready content, and a local mentions list. Then, at 60 days, we rerun the identical queries and send you the before/after.
$299, one-time. No subscription. We guarantee the work and the retest — and we’re honest that nobody can guarantee an AI’s answer.
See the AI Visibility PackageNot sure where you stand? Start with the $249 Visibility Audit — it includes the AI answer check for your market, and it credits in full toward the package.