All Posts
AI Search / GEO

How AI Engines Choose Local Businesses (And How to Be the One They Name)

Joel House, Founder, Xpand Digital
Joel HouseForbes Agency Council
Founder, Xpand DigitalJuly 12, 202613 min read

Someone types “best family lawyer in Denver” into ChatGPT and a name comes back. It isn't yours. That gap is a business problem, and it's a solvable one.

I run an agency that instruments this for a living, so I get the question constantly: how does ChatGPT recommend local businesses, and why does it keep naming the same three competitors instead of me? The honest answer surprises people. ChatGPT does not have a private, pre-baked ranking of the plumbers or dentists in your city. There's no leaderboard sitting on a server somewhere with your business at position #14. When someone asks for a local recommendation, the engine assemblesone on the spot — in a second or two — from whatever the open web can tell it about your category in that location.

That's the whole game, and it's good news if you know what you're doing. It means the recommendation is made from a specific, knowable set of inputs. Control the inputs and you change the output. This guide is the operator's view of exactly which inputs matter, in what order, and what to do about each one — written from running these campaigns, not from theory.

First, the scale of the opportunity. MentionLayer's Q1 2026 AI Visibility Indexanalyzed 95,392 data points across 1,004 businesses and found 65.9% of them were effectively invisible in AI search — not named, not cited, not surfaced. Most of those businesses have a live website and a Google listing. They simply never handed the engines a clean signal to pull. The bar to be the named business is lower than it looks, because two-thirds of your market hasn't shown up to the table yet.

What actually happens when someone asks for a local recommendation

Before the inputs, it helps to see the sequence. When a prospect asks ChatGPT, Perplexity, or Google's AI Overviews for a local business, roughly five things happen in the couple of seconds before an answer appears:

  1. It parses intent and category. “Good SEO agency,” “emergency plumber,” “family lawyer” — the engine identifies what kind of business you want.
  2. It resolves the location. From the prompt (“in Austin”), from prior conversation, or from context. If it can't pin a location, the answer gets generic and national.
  3. It retrieves live sources. The model queries its search layer — for ChatGPT that's Bing's index plus map data — and pulls back a working set of pages, profiles, and listings.
  4. It synthesizes a short list. It reads the retrieved sources, cross-references who appears repeatedly, and drafts a recommendation of usually three to five businesses.
  5. It attaches citations. It links or names the sources it leaned on, which is why directory pages and “best of” roundups show up under AI answers so often.

Every one of those steps is a place you either make the shortlist or fall off it. The businesses that get named aren't the ones with the biggest ad budget — there is no ad slot inside that answer. They're the ones whose evidence is clean, consistent, and repeated across the exact sources the engine reaches for. Below are the five inputs that decide it, ranked by how much leverage each one carries.

The five inputs that decide who gets recommended

Think of these as the evidence file the engine builds about your business in real time. Weak on all five and you're part of the invisible 65.9%. Strong on the first three and you start getting named in your category. Strong on all five and you become the default answer.

01

Structured local data — can the engine confirm who and where you are?

This is the foundation, and it's where most invisibility actually starts. Before an engine will name you, it needs to be confident about three boring facts: what you do, where you operate, and that you're a real, active business. That confidence comes from structured data — your Google Business Profile, your category selection, and a name, address, and phone number (NAP) that read identically everywhere they appear. When your website says “Suite 200,” your Business Profile says “Ste. 200,” and an old directory says “#200,” the engine sees ambiguity, and ambiguity gets you dropped from a synthesized answer where the model is trying not to be wrong.
What the operator does
Claim and fully complete the Google Business Profile — correct primary category, service areas, hours, services list. Audit NAP consistency across your site, Business Profile, and the top directories in your industry, and fix every mismatch to one canonical version. Add LocalBusiness structured data (schema) to your site so the facts are machine-readable, not just human-readable.
02

Third-party consensus — does the open web already agree you're good?

Engines are built to avoid recommending something they can't back up. So the single strongest positive signal is consensus: independent sources naming you, at volume, without you having to claim it yourself. Reviews are the fastest form of this — a profile with 180 recent reviews at 4.7 stars reads as a live, trusted business the model can safely put its name behind. “Best [category] in [city]” roundups are the second form, which is exactly why those listicles get cited under AI answers so often: the engine is borrowing someone else's editorial judgment. Industry directories the model already trusts are the third. When three independent sources say the same thing about you, the engine treats it as fact.
What the operator does
Run a real review engine — steady volume and recency beat a one-time push, because both count. Get honestly included in the “best of” roundups and directories that already rank for your category and city; those are the pages engines retrieve first. Earn editorial mentions through digital PR so your name appears in content you didn't write. This is the slow, compounding work, and it's the moat once it's built.
03

Entity clarity — does the model understand you as a distinct thing?

An “entity” is how a machine holds the concept of your business — a specific company, in a specific place, that does a specific thing, connected to a founder, reviews, and a category. When your entity is clear, the engine can retrieve you with confidence and match you against the right queries. When it's muddy — inconsistent naming, no clear founder or team, a homepage that could describe forty different companies — the model can't hold you steady, and it defaults to competitors it understands better. This is the input almost nobody optimizes, which is precisely why it's an edge.
What the operator does
State plainly, in plain text on the page, what you do and where. Give the business a real named founder or team with a consistent bio across the site and third-party profiles — a recognizable person hardens the entity. Keep the business name spelled and formatted one way everywhere. Use Organization and Person schema to spell the relationships out for the machine.
04

Citable content — do your pages answer the exact question?

Retrieval engines don't reward the longest page; they reward the page they can quote. When someone asks “who does SEO for law firms in Los Angeles,” the engine wants a page that says, in extractable sentences, that you do SEO for law firms in Los Angeles — who it's for, what it costs, what the outcome looks like. A page buried in adjectives and stock phrases gives the model nothing clean to lift, so it lifts a competitor's clearer sentence instead. This is where classic content marketing and AI visibility diverge: you're not writing to impress a reader for ninety seconds, you're writing sentences a machine can pull and stand behind.
What the operator does
Build service and location pages that answer the real questions directly — category, city, price range, who it's for — in the first hundred words, not the last. Write in plain declarative sentences the engine can extract. Add an FAQ that mirrors how people actually phrase the ask. Our city pages, like AI SEO in Los Angeles, are built exactly this way — to be quotable, not just readable.
05

Freshness and activity — does the business look alive right now?

Engines discount businesses that look dormant. A Business Profile last updated eighteen months ago, a blog that stopped in 2024, reviews that dried up — each one nudges the model toward a competitor that looks active this quarter. Recency is a proxy for “still in business and still good,” and the engine uses it to hedge against recommending somewhere that closed. This input carries less weight than the first four, but it's the cheapest to fix and it breaks ties.
What the operator does
Keep the Business Profile moving — posts, photos, updated hours, fresh Q&A. Maintain a steady trickle of new reviews rather than a single burst. Publish and update pages on a real cadence. None of this is heavy; it just has to be ongoing, because “alive right now” is a signal you can only send continuously.

“The engine isn't choosing the best business. It's choosing the business with the cleanest, most-repeated evidence. Those are different things — and the second one is winnable.”

Joel House · Founder, Xpand Digital

Why most local businesses never get named

The invisible 65.9% isn't invisible because their work is bad. It's invisible because of a handful of repeatable failures I see on nearly every audit. None of them are exotic. All of them are inputs the business controls.

  • The location never resolves. No LocalBusiness schema, a thin or unclaimed Business Profile, a homepage that never states the city plainly — so the engine can't confidently place them, and defaults to national brands.
  • No third-party consensus. A handful of old reviews, no directory presence, no inclusion in a single “best of” roundup. The engine has nobody but the business vouching for the business, and that isn't enough to name it.
  • The pages are unquotable. Beautiful, brand-forward copy that says everything except what they do, for whom, and where. The model finds nothing clean to lift.
  • The entity is muddy. The business trades under three slightly different names, has no visible team, and reads like a template. The model can't hold it steady enough to retrieve.
  • Nobody is measuring it. The business has no idea it's absent, because it's still watching Google rankings and never once asked ChatGPT what it recommends. You can't fix a gap you can't see.

Notice what's noton that list: domain age, ad spend, agency size, or how long you've been in business. The AI answer is assembled from evidence, and evidence is something you can build in months, not years. That's the structural reason a sharp local operator can leapfrog an incumbent in AI recommendations far faster than they ever could in classic search.

The operator's checklist: becoming the named business

Here's the sequence I'd run for any local business that wants to start getting recommended, ordered so each step compounds on the last. It maps directly onto the five inputs above.

  1. Fix the foundation first. Claim and complete the Business Profile, reconcile NAP to one canonical version everywhere, and add LocalBusiness schema. Nothing downstream works until the engine can confirm who and where you are.
  2. Turn on a review engine. Systematic, ongoing review generation — not a one-time ask. Volume and recency are the fastest consensus signal you can build, and they seed the language the engine matches queries against.
  3. Get into the roundups and directories. Earn honest inclusion in the “best of” lists and industry directories already ranking for your city and category — those are the exact pages engines retrieve and cite.
  4. Rewrite your pages to be quotable. State the category, city, who it's for, and price range in plain sentences up top. Add an FAQ that mirrors real phrasing. Make it easy for a machine to lift a sentence and stand behind it.
  5. Harden your entity. One business name, one founder bio, consistent everywhere, with Organization and Person schema wiring the relationships together.
  6. Keep it alive. Ongoing profile posts, a steady review trickle, a real publishing cadence. Freshness breaks ties in your favor.
  7. Measure share of answer. Prompt the engines monthly for your core queries and log named / cited / absent. That trend is your real scoreboard.

This is the exact work our GEO agency program runs, and it's the reason I'll say something most agencies can't: we don't guess at these inputs. We built our own AI-visibility platform, MentionLayer, and our own digital-PR engine, PressForge, and I wrote AI for Revenueon the strategy behind it. When we tell you which sources ChatGPT reached for on a query, it's because we instrumented it — not because we read a blog post. An agency selling GEO without its own measurement is optimizing in the dark.

ChatGPT vs. Perplexity vs. Google's AI Overviews

The five inputs win everywhere, but the engines weight their sources differently, and it's worth knowing where the emphasis shifts:

  • ChatGPT and SearchGPT lean on Bing's index and map data. Business Profile completeness and presence in the directories Bing surfaces carry real weight here. If you've only ever optimized for Google, this is where you're most likely under-indexed.
  • Perplexity is citation-first — it rewards content it can quote and link directly, and it shows its sources aggressively. Quotable pages and clean editorial mentions punch above their weight on Perplexity.
  • Google's AI Overviews sit on top of Google's own local stack, so Business Profile signals, Maps activity, and review volume matter most. Strong classic local SEO feeds AI Overviews the most directly of the three.

The strategic point: you optimize the inputs, not a single engine. Do the foundation, consensus, entity, and content work well and you show up across all three, because they're all reading versions of the same open-web evidence. Chase one engine's quirks and you win narrowly and briefly. This is also why AI recommendations and traditional local search reinforce each other — the profile and citation work that feeds local SEO is the same work that feeds the AI answer, and neither replaces the other.

How to know if it's working

Classic rank tracking will tell you nothing about this. Your position #3 in Google says nothing about whether ChatGPT names you, because the AI answer is synthesized from a different, narrower slice of the web. You need a different scoreboard, and it's simple to keep.

Pick ten to twenty prompts your customers would actually type — “best [category] in [city],” “who should I call for [problem] near me,” “affordable [service] in [city].” Run them across ChatGPT, Perplexity, and Gemini once a month, and log one of three outcomes for each: named (you're in the recommendation), cited (a source about you was used), or absent. That share-of-answer, tracked over time, is the metric that matters. When the foundation and consensus work lands, you watch queries move from absent to cited to named — usually in that order, over a few months.

The businesses that win the AI recommendation aren't the ones with the flashiest site. They're the ones that treated the engine's five inputs as a checklist and worked it methodically while their competitors argued about whether AI search mattered yet. It matters now. Two-thirds of the market is still invisible — which means the seat next to the named businesses is open, and it goes to whoever shows the engine the cleanest evidence first.

Getting named by AI engines is one layer of a full local growth program. These pair directly with the work above:

  • GEO agency services— how we run generative engine optimization end-to-end, instrumented with our own tooling across every city we work.
  • Local SEO services— the Business Profile, citation, and review foundation that feeds both classic local rankings and AI recommendations.
  • ChatGPT SEO— the engine-specific playbook for showing up in ChatGPT and SearchGPT answers.
  • Free AI visibility audit— we'll prompt the engines the way your customers do and show you exactly where you're named, cited, or absent.

Frequently Asked Questions

Joel House, Founder, Xpand Digital
Founder, Xpand DigitalJuly 12, 2026

Find out what the engines
say about you

We'll prompt ChatGPT, Perplexity, and Gemini the way your customers do, show you where you're named, cited, or absent, and map the exact inputs to fix first. No pitch deck — just the evidence file.

Get a Free AI Visibility Audit See How GEO Works
Put it to work

Work with the team that wrote the playbook.

Where to start