“ChatGPT is recommending my business. How did that happen?”

A buyer calls and says, “ChatGPT told me to contact you.” You did not pay for that. You did not submit anything. Somewhere between the buyer’s question and the answer, a system decided your business belonged in the reply. This article explains what that system did, which of your signals it read, and how to make it happen again for the next question, or for the first time if it has not happened yet.

We work from what we observe when we put real buyer questions to the engines on behalf of clients. Names and specifics stay with the clients; the patterns are general.

What ChatGPT does with a buying question

When someone asks ChatGPT “who does commercial snow removal in Mississauga” with search enabled, four things happen in about ten seconds:

It rewrites the question into one or more search queries and runs them against a web index. For ChatGPT that index is Bing’s, supplemented by OpenAI’s own crawler.
It reads a handful of results, usually five to ten pages: company sites, directories, review platforms, local news, association member lists.
It cross-checks the businesses it finds against structured listing data. In our sampling, ChatGPT often renders a business card with an address, category, hours and star rating drawn from map and directory listings.
It writes one answer naming two to five businesses, with citations back to the pages it read.

If search is not enabled, or the question is general enough that it does not trigger a search, the model answers from memory. For most small and mid-sized businesses that memory is empty. Being named without search means you have been mentioned widely enough, for long enough, to end up in training data. That is a multi-year project, not a quarter’s.

So a recommendation with search on is the realistic target, and it has five moving parts.

The five signals behind a recommendation

Signal	What it looks like	How to check
Findable	A page of yours (or about you) ranks in Bing for the question or a close variant	Search Bing for the buyer question; are you in the top 10?
Answer-shaped	The page states the service, the area, the price range or process in plain HTML text	Open the page with JavaScript off; is the fact still there?
Corroborated	A directory, association, news site or partner page says the same thing about you	Search your business name plus your city; what appears besides your site?
Consistent	Name, address, phone, category and service area agree across site, Google Business Profile, Bing Places and directories	Compare the four side by side
Unambiguous	Your name is not shared with another business the model might confuse you with	Ask ChatGPT “tell me about [your name]”; does it describe you?

Most businesses that are already being recommended score well on three of the five without having tried. The ones that are missing usually fail one specific signal, and the failure is visible in the sources ChatGPT cites instead of them.

Signal one: findable

ChatGPT does not have its own copy of the web. When it searches, it is searching Bing. A site that has never been submitted to Bing Webmaster Tools, or that Bing has crawled thinly, is invisible to ChatGPT search no matter how well it does on Google.

Two fixes, both free:

Verify the site in Bing Webmaster Tools and submit the sitemap. You can import the verification from Google Search Console.
Turn on IndexNow if your platform supports it. It pushes new and changed URLs to Bing immediately instead of waiting for a crawl.

Then check robots.txt allows OAI-SearchBot (the crawler behind ChatGPT search) and ChatGPT-User (the fetch that happens when a user clicks a citation). Blocking either one, often through a default firewall rule, removes you from the pool.

Signal two: answer-shaped

The engine reads your page the way a hurried assistant would: looking for the sentence that answers the question. “We provide commercial snow removal for property managers across Mississauga and Oakville, with 24-hour response and seasonal contracts from November to April” is an answer. “Winter solutions for your peace of mind” is not.

For each question you want to be recommended for, there should be one page where the answer is stated in the first screen, in text, with the facts an engine can quote: what, for whom, where, how much or how it is priced, how long it takes. Structured data (Service, LocalBusiness, FAQPage) should repeat the same facts so nothing is left to interpretation.

Signal three: corroborated

This is the signal businesses most often underestimate. In our sampling, for the same buyer question, engines cite industry associations, local news and directory pages at least as often as company websites, and sometimes instead of them. A claim on your own site is a claim. The same claim on an association member page or in a local newspaper is evidence.

Cheap sources of corroboration, in rough order of effort:

Your industry association’s member directory, with a full profile.
Chamber of commerce and regional business directories.
Supplier and partner pages that list who they work with.
One local news mention, even a small one: a project, a hire, a donation.
Review platforms where your category is active.

The pattern to notice is that these are pages someone else controls. That is exactly why engines trust them.

Signal four: consistent

ChatGPT’s business card comes from listing data. If your Google Business Profile says one address, your site says another, and a directory has your old phone number, the engine either hedges or drops you. Consistency is boring and it is decisive.

Check name, address, phone, website, category and service area across your site footer, structured data, Google Business Profile, Bing Places and the two or three directories that matter in your industry. Make them identical, down to the suite number and the spelling of the street.

Ratings matter here too. A card showing a low score with a handful of reviews is a weak recommendation even when the engine makes it. Each new review moves the number more when the count is small.

Signal five: unambiguous

We have watched an engine describe a client by reciting facts about a different company in another country that happened to share two words of its name. The answer was confident, cited, and wrong. Nothing on the client’s site was at fault; the model simply merged two entities.

The fix is to give the engine enough to tell you apart:

Use your full legal name and your location together in structured data (name, legalName, address) and list every spelling in alternateName.
Put a plain sentence on your About page: who you are, where you are, since when.
Say the same in llms.txt.
Do not name the other business anywhere on your site; that only strengthens the association.

How to check yourself, honestly

You can get a first reading without any tool: ask ChatGPT, with search enabled, a few questions your buyers would actually ask, and note who it names and which sources it cites. Two cautions. Keep your own name out of the question; if you ask “is Northline Fabrication good,” the model will mention Northline Fabrication whether or not it knows anything. And do not read a single answer as a verdict. The same question on the same day can cite you once and skip you once. Only what repeats across several tries is worth acting on.

If you are not being named

Work the five signals in the order above. Findability first, because nothing else matters if Bing cannot see you. Then one answer-shaped page per question. Then two or three corroborating pages. Then the consistency sweep. Then the disambiguation, if your name is common.

Expect the first movement in ChatGPT’s answers three to eight weeks after the changes are indexed. Gemini and Google’s AI Overviews follow organic ranking more closely and can move faster or slower depending on how competitive the question is.

Where to start

If you would rather see the current picture before you start, we run the ten-question exercise above at scale, twenty real buyer questions across Gemini and ChatGPT, and send back a PDF showing who each engine cited instead of you. It is free and takes two business days.

Questions people ask

Can I pay to be recommended by ChatGPT?

No. There is no advertising placement inside ChatGPT answers. Recommendations come from what its web search retrieves and what the model has learned. The levers are the same as being cited anywhere: be findable, be factual, be consistent, be mentioned by others.

Why does ChatGPT recommend a competitor and not me?

Usually because the competitor has a page that answers the exact question in plain text, is indexed by Bing, and is corroborated by a directory, association or news source. Ask the same question yourself with search enabled and look at the sources it cites; the gap is usually visible in one screen.

How do I get my business listed on ChatGPT?

There is no listing to submit. Make sure Bing has indexed your site (Bing Webmaster Tools), your Google Business Profile and Bing Places are complete and match your site, your key pages state services, area and pricing in HTML text, and at least one third-party page describes you. Then ask the buyer questions yourself and check.

Why does ChatGPT describe my business incorrectly?

Either it read a stale or thin page, or it merged you with another business that shares your name. Fix the first by updating the page it cited. Fix the second by stating your full legal name, location and aliases consistently in your structured data, your About page and your llms.txt.

( WHERE YOU STAND )

Does AI recommend you?

We put twenty real buyer questions about your category and market to Gemini and ChatGPT and send you a PDF showing who they cite instead of you, plus fourteen technical checks. Free, no call required.

Request the free audit