How ChatGPT picks businesses to recommend

People now ask ChatGPT the question they used to type into Google: who around here is good at this. The adoption is not in doubt: Pew Research Center put the share of US adults who have used an AI chatbot at 49% in February 2026, up from 33% two years earlier, with 42% using one to search for information. What nobody neutral has yet measured is how often those questions are about a local business specifically, so treat the size of this channel as unknown and the direction as obvious. It is early, messy, and full of people selling a product called “LLM SEO” that looks a lot like last year’s content package with the label peeled off.

We know the channel works, because that is how one of our own clients found us. He did not come in through an ad. He asked ChatGPT, and Vertana was the answer it gave. That is the whole promise of this page, and it is also the limit: we cannot make the model do that on demand, for you, next month. We can make you the kind of business it has something true to say about.

What LLM SEO actually is

LLM SEO is the work of becoming a source an assistant can quote without inventing you. The model does not have a special index of “approved small businesses.” It has training data, it has whatever it can retrieve, and it has a bias toward pages that state facts in plain language rather than marketing copy.

You will also see this work sold as answer engine optimization or generative engine optimization. The acronyms differ. The tasks underneath them are close to identical, and none of them are a product you should be billed for separately. If you are being quoted for one of them, the side-by-side is in SEO vs AEO vs GEO.

In practice that means:

• Pages that name the service and the town in the first screen, not after a hero video.
• One topic per page, so the model can tell a sports-massage page from a couples-massage page.
• Name, address, phone, hours, and services that match your Google Business Profile and the directories that already list you.
• Structured data that repeats those same facts, not a second set of keywords stuffed into JSON-LD.
• Other places on the web that say the same thing: reviews, citations, a LinkedIn or association page that is actually yours.

None of that is new. It is the same work as monthly SEO, described from the assistant’s side of the table. If a page is too thin for Google, it is too thin for ChatGPT.

What ChatGPT can see, and what it cannot

Treat the assistant like a hurried researcher, not like a customer who will click around. It will not watch your video. It will not fill in your contact form to see what happens. It will not infer that you “serve the whole metro” from a vibe. If the page does not say it, the model has to guess, and guessing is how you get recommended for the wrong thing, or not at all.

It also cannot see inside your booking software, your private Facebook group, or a site that only paints the words in after JavaScript runs. Content that exists in the HTML as delivered is content it can use. Content assembled in the browser is a coin flip.

Reviews matter here the same way they matter on the map. An assistant asked for “a good plumber in Lakewood” will lean on corroboration: other people saying you exist, in the place you claim, for the work you claim. A five-year-old site with three reviews and a homepage that says “we care about quality” is not quotable. A three-page site that names the services, the towns, and the license, with a matching profile, often is.

What to fix this week

If you want a first pass you can do without hiring anyone, do it in this order. It is sorted by how often each gap is the reason an assistant has nothing useful to say.

1. Search your business name in ChatGPT, Gemini, and Google. Write down what each one gets wrong: the town, the services, a competitor it names instead of you, or a version of you that closed two addresses ago.
2. Put the facts on the homepage in a form a stranger could copy: what you are, where you work, who you work with, how to book. Not a slogan. A description.
3. Give each service you actually sell its own page. Five services on one page is one weak answer.
4. Make the Google Business Profile, the website, and the top directories agree on name, address, phone, and categories. Mismatches are how models pick the wrong listing.
5. Ask the next ten happy customers for a review that names the work you did, not just a star rating.

If you cannot find yourself at all in Google, stop here and work through why the business is not showing up first. An assistant cannot recommend a business the rest of the web has not noticed.

When this is the wrong spend

Skip the LLM-SEO package, including ours framed that way, if any of these are true:

• You do not have a website a crawler can read, or the one you have is a locked template nobody can edit. Fix the site first.
• The Google Business Profile is unclaimed, suspended, or in the wrong category. That is a free afternoon, not a strategy.
• Nobody in your trade searches or asks an assistant. A business that lives entirely on one repeat contract does not have an AI-search problem.
• You need leads this month. That is ads, or the phone, not a model that updates on its own schedule.

The honest offer is the other way around. Do the SEO. Build pages worth quoting. Keep the profile accurate. The assistant recommendations, when they come, ride on that work. They are not a separate product, and they are not a guarantee.

Common questions

Can you pay to rank in ChatGPT?

No. There is no ads product that puts your business into a ChatGPT answer, and anyone selling a guaranteed listing is selling something else. What you can do is make the facts about your business easy to find, easy to quote, and easy to corroborate on the rest of the web.

Is LLM SEO different from regular SEO?

The vocabulary is different. The work is mostly the same. Assistants read pages, structured data, directories, and reviews. A site that states clearly what you do and where you do it, with a Google Business Profile that matches, is doing LLM SEO whether or not anyone called it that.

How long does it take to show up in ChatGPT?

There is no reliable timeline, because no one controls the model. Some businesses are already named because they have been written about clearly for years. A new site with thin pages should not expect to be recommended next week. Treat this as a side effect of doing the SEO properly, not as a campaign with a due date.

Do I need special schema for AI search?

You need structured data that matches what a person can read on the page: your name, services, area, hours. Invented markup, FAQ schema with no FAQ, and keyword stuffing in JSON-LD do not help, and they can hurt. If the visible page is thin, schema will not save it.

If the underlying job is getting found on Google and in answers like these, that is the monthly SEO work. Local map-pack work sits next to it on local SEO services. Named proof of the same stack is on law and wellness.