TL;DR
How does AI search pick which local businesses to recommend? It pulls live data from the web and directories, then recommends the businesses whose information is the most consistent, corroborated, and machine-readable.
- Fix three things in order: matching name, address and phone everywhere, a complete Google Business Profile, and a review somewhere besides Google.
- Audit every place your business appears online before adding any new tool.
- Consistency decays as hours and listings change, so automate the checks.
Here's the direct answer, because I know that's what you came here for: ChatGPT, Gemini, and Google's AI Overviews aren't ranking your business against a fixed list the way old school SEO worked, they're retrieving live information from the web and directories in real time and then writing an answer, and they cite whoever is easiest to verify and easiest to pull clean information from. That means the businesses showing up aren't necessarily the best ones, they're the ones whose data is consistent, corroborated, and structured in a way a machine can actually parse without guessing. So if your competitor is getting recommended and you're not, it's rarely because they're better, it's because their information is easier for the machine to trust.
That's actually good news, because it means this is fixable, and it doesn't require you to be a tech person or hire a full marketing department. It requires the same thing I tell every client who thinks they need more software: you probably don't need a new tool, you need your existing information straightened out and kept that way. Let's walk through how these things actually decide who to recommend, then get into the specific fixes in order.
How AI Tools Actually Decide Who to Recommend
When someone asks ChatGPT or Gemini for a plumber, an accountant, or a staffing agency near them, the tool is pulling from a mix of what it was trained on and what it can retrieve live from the web at the moment of the question, and then it's synthesizing that into a conversational answer with sources attached. There's no dashboard where you submit your business and wait for approval, and anyone selling you that is selling you something that doesn't exist. Google's AI Overviews work similarly, and they've grown fast, showing up in roughly 6.5% of tracked searches back in January 2025, spiking near 24.6% by July, and settling around 15.7% by November, which tells you this isn't a fad you can wait out (source). For local, informational searches specifically, a Whitespark case study found AI Overviews showed up in 92% of results (source), and ChatGPT alone had roughly 800 million weekly users as of October 2025 (source). So the volume of people asking AI where to go instead of googling it themselves is already massive, and it's not slowing down.
The Signals That Actually Move the Needle
Across the different platforms, the same handful of signals keep showing up as the ones that matter, and none of them are exotic. One breakdown of how these models weight information found that ChatGPT puts roughly 20% weight on review quality and sentiment and 15% on location, while Gemini puts about 22% on Google Business Profile quality and another 20% on review volume, rating, and content (source). And looking at where these engines are actually pulling their information from, research on professional services found that 80.9% of the listicles AI engines cited came from third-party sources rather than the business's own website (source), which means your own site copy matters less than what other people and directories are saying about you. That lines up with what I see doing audits for clients, the businesses that get found are the ones whose name, address, and phone number match everywhere, whose reviews are spread across more than just Google, and who have at least a few outside mentions vouching for them.
What's Hype and What's Real
Structured data (schema markup, the code that tells a machine "this is a business, here's the phone number, here's the hours") genuinely helps, but it's not a magic bullet, and I get frustrated watching other consultants sell it like one. Google's own people are split on how much it matters for ranking specifically. Ryan Levering has said structured data helps their systems, while John Mueller has said it doesn't directly influence ranking but does support engagement and semantic understanding (source). So schema is worth doing, but it's one piece of a system, not the whole system, and if someone tells you adding a tag will get you into ChatGPT's answers by itself, they're overselling it. This is the same problem I see with most AI consultants generally, they chase the shiny technical fix instead of doing the unglamorous work of getting the underlying information straight first.
The Fix, Step by Step
Before you touch anything, you need to know where your business information currently lives, which means pulling up your website, your Google Business Profile, your top directory listings (Yelp, BBB, any industry-specific ones), and your review platforms, all in one place so you can compare them side by side. Skip this step and you'll end up fixing things that were already fine while missing the actual gaps, which is the most common mistake I see.
- Do a structured audit of where your business data actually lives. List every place your name, address, phone number, hours, and services appear, including old directory listings you forgot existed. You'll know it worked when you have a single sheet showing every source side by side. The mistake here is assuming your website is the source of truth, when for AI engines it's usually the least trusted one.
- Fix every inconsistency in your name, address, and phone number. If your website says "Suite 200" and your Google Business Profile says "Ste 200B," fix it everywhere until it matches exactly. You'll know it worked when a directory search on your business name pulls the same details everywhere. The mistake is fixing it in one place and assuming it propagates, it doesn't, you have to go update each listing manually or through an aggregator.
- Fill out your Google Business Profile completely, hours, services, photos, all of it, because Gemini weights this heavily. You'll know it worked when there are no "incomplete" prompts left in your profile dashboard. The mistake is filling it out once at setup and never touching it again as your services or hours change.
- Get reviews on at least two platforms beyond Google. Reviews are one of the strongest signals across every model, so a business with only five Google reviews and nothing else is easier to skip than one with reviews on Google, Yelp, and an industry directory. The mistake is only asking for reviews right after a good outcome and ignoring the slow drip that keeps volume growing.
- Add plain-text content that answers specific customer questions, not marketing copy, actual questions people ask you on the phone or in emails, written out in normal sentences on your site where a machine can read the text (not buried in an image or PDF). You'll know it worked when your FAQ section reads like something a real customer would ask, not like ad copy.
- Add schema markup as a supporting layer, not the whole plan. LocalBusiness and FAQPage schema help machines parse your page faster and with more confidence, so it's worth doing, but treat it as one piece among several.
- Get at least one third-party mention, a press writeup, an industry guide, a directory feature, anything outside your own site vouching for you, since these are cited far more often than a business's own pages.
Doing this well usually means unifying your data sources for better AI discoverability, because right now that information is probably scattered across five different logins that nobody checks consistently, and that scattering is the actual bottleneck, not a lack of tools.
Why This Breaks Again in Six Months
Here's the part almost nobody tells you: this isn't a project you finish, it's a system you maintain. Your hours change, a new employee stops updating one directory, a review platform adds a new field, and six months later your data is inconsistent again without you noticing, and you're back to being invisible in the same answers you just fixed. This is exactly the same "leaks in the intake process" problem I deal with for clients on their day-to-day operations, and the fix is the same too, you automate the maintenance and consistency that keeps you visible instead of manually re-checking everything on a calendar reminder you'll eventually ignore. From a one-person shop to a 500-person operation, the mechanics of how AI visibility scales from solo operators to larger teams are the same, it's just a matter of how much data you're keeping in sync and how much of that syncing gets automated versus done by hand.
Score Yourself: A Quick AI Search Readiness Check
You can score your own readiness with seven yes or no questions: do you have schema installed and validated, is your NAP identical everywhere, is your Google Business Profile fully filled out, do you have reviews on two or more platforms beyond Google, does your site answer specific questions in plain text, have you checked how AI describes your business in the last three months, and do you have at least one third-party mention. Count your yeses out of seven. Zero to two means you're essentially invisible to AI search right now and should start with NAP consistency, a complete Google Business Profile, and plain-text FAQ content, in that order. Three to five means you're partially visible and just need to knock out whichever specific items you answered no to. Six or seven means you're in good shape and the remaining work is maintenance, not a rebuild.
Say, hypothetically, you run a small landscaping company and you check your own answers: schema, no. NAP consistent, mostly, except one old directory listing from years ago. Google Business Profile, yes, fully filled out. Reviews on two platforms, no, just Google. Plain-text FAQ content, no, your site is mostly photos. Checked how AI describes you, no. Third-party mention, no. That's two yeses out of seven, which puts you in the invisible category, and your first three moves are obvious: fix that old directory listing, get reviews going on one more platform, and write out plain-text answers to the five questions your customers ask you most often.
Questions people ask
How do I know it's actually making me money?
I won't tell you there's a guaranteed revenue bump from any of this, because there isn't one, results depend on whether you actually follow through and keep the information current. What I can tell you is that AI-driven search is already a large enough slice of how people find businesses that not showing up is a real cost, and fixing the underlying data is cheap compared to what you lose sitting invisible.
How long does it take to get going?
The audit itself, figuring out where your data lives and what's inconsistent, can usually happen in a matter of days. Getting everything corrected and the first round of content and reviews in place takes longer, often a couple of weeks depending on how scattered things were to start, and then it becomes an ongoing maintenance habit rather than a one-time project.
I looked into hiring an automation consultant and it seems too expensive or complicated for someone like me.
It might well be, if someone's trying to sell you a full custom platform when what you actually need is a data cleanup and a handful of automated checks. That's why I always scope this down to exactly what a business needs and can afford, because overbuilding something you won't use or can't pay for doesn't help either of us.
If you want a clear picture of where your own business data is inconsistent and what's actually worth fixing first, that's exactly what a structured audit of where your business data actually lives is for, and it's the same starting point I use with every client before touching a single tool. Start there, find the gaps, and fix them in order instead of guessing.
