AI visibility for small business: how to find out if ChatGPT recommends you

Here is the short answer. To find out whether ChatGPT, Gemini, Perplexity, and the others recommend your business, you run the exact questions your buyers ask, across each engine, several times, and you write down who gets named and which sources the answer leaned on. Where a competitor shows up and you do not, the reason is almost always the same: they are present in the places the engine reads and you are not. The fix is to get present in those places. You can do a rough version of this yourself for free in an afternoon, and a complete version for the price of a single service call.

The rest of this page is the long answer: why small businesses are more exposed than big ones, what actually moves an AI recommendation for a local or niche business, how to measure it without a marketing budget, and a worked example of what a scan turns up for a real-looking business.

Why a small business is more exposed than a big brand

People used to type "plumber near me" into Google, get a map with ten pins, and pick one. Now a growing share of those same people open an assistant and type something closer to how they actually think: "my kitchen sink is leaking under the cabinet, who is a good plumber in Leeds and roughly what will it cost." The assistant does not return ten pins. It returns a short paragraph that names three to five businesses and says a sentence about each. That paragraph is the entire shortlist. If your name is not in it, you were never in the running, and the buyer has no idea you exist.

This hurts a small business more than a chain for a simple reason. A national brand is mentioned in thousands of places, so an engine has seen its name constantly and will surface it even when its own pages are thin. A small business lives or dies on a much smaller set of signals: a handful of directory listings, its reviews, the odd local article, a Reddit thread or two. When those are missing or inconsistent, there is almost nothing for the engine to grab, and you fall out of the answer entirely. You do not get a low ranking. You get left out of the sentence.

It is worth being clear about what does not save you here. Your Google Business Profile controls how you appear in Google Maps and the local pack. It does not control what ChatGPT says when someone asks it for a recommendation, because ChatGPT is not reading your Business Profile the way Google Search does. A perfectly optimized profile can sit next to an assistant that has never heard of you. These are two different systems, and winning one does not win the other.

What actually moves an AI recommendation for a small business

The good news is that the levers are concrete and mostly free. An assistant names the businesses it can find and quote. So the whole game is being findable and quotable in the specific places it looks. For a small business, that means four things.

• Presence in the sources engines actually read. When an assistant answers a "best X near me" question, it leans on review sites and directories it can quote, on local "best of" roundup articles, and on threads where real people ask for recommendations, Reddit chief among them. If your business is listed, reviewed, and discussed in those places, the engine sees you. If you are absent, it cannot name what it cannot find.
• Pages that answer real buyer questions in plain language. A page titled for the exact thing a buyer asks, written the way they would say it, gives the engine a clean sentence to lift. "Emergency boiler repair in Bristol, same-day, typical cost" is quotable. A homepage that says "your trusted local partner for heating solutions" is not. Write the page a buyer would actually search for and answer the question near the top.
• Consistent name, address, phone, and structured data. If your business name, address, and phone number read the same across your site, your directory listings, and your profiles, an engine trusts that you are one real business. If they conflict, it hedges and may skip you. Adding basic structured data (LocalBusiness schema) and a plain, honest pricing or services page makes you easy to parse and easy to summarize.
• Earned mentions. Every time an independent source names you, a local news piece, a supplier's partner list, a happy customer's Reddit reply, the engine has one more reason to surface you. Repetition across independent places reads as consensus. You do not need hundreds of these. For a local business, a handful of genuine mentions moves the needle.

Notice what is not on the list: paying the model, being the biggest, or stuffing hidden instructions into your page. There is no ad slot inside the answer, and hidden text telling an assistant to recommend you gets ignored. This is earned presence, and a small business can earn it without a large budget because the sources that matter for local and niche categories are shallow enough to actually influence.

What you can control and what you cannot

It helps to separate the parts of this you own from the parts you only influence. Spend your limited time on the first column.

The pattern is clear. You cannot dictate the answer, but you can stack the inputs so heavily in your favor that being named becomes the likely outcome across most runs. That is the realistic goal: not to win every single answer, which no one does, but to be present enough that you show up in the majority of them.

Measuring it cheaply: a one-time scan versus ongoing tracking

You cannot fix what you cannot see, and guessing which sources matter for your specific trade and town wastes months. There are two honest ways to measure, and they suit different moments.

A one-time Scan answers the question "where do I stand right now, and what do I fix first." Saymetry's Scan is $249. It takes the real questions your buyers ask, runs them across seven AI engines, ChatGPT, Claude, Gemini, Perplexity, Grok, DeepSeek, and Mistral, audits your website, and returns a report with an action plan within the hour. That is the right choice for most small businesses starting out. You get a clear picture and a prioritized list, once, for less than the cost of a paid ad campaign that would not touch this problem anyway.

Ongoing Tracking answers a different question: "is what I am doing working, and am I holding my ground as answers shift." Because AI answers change often and without warning, a business that depends on these recommendations benefits from watching the trend rather than checking once. Tracking is $149 a month, or $119 a month billed yearly. It reruns your questions on a schedule so you can see your presence rise or slip and catch it when a competitor makes a move. Most owners start with the Scan and move to Tracking only once they have proven the fixes are worth defending.

A quick note on scale, so you know the ground is real. In Saymetry's own scan of the AI-visibility category, 60 buyer questions were run across 7 engines (ChatGPT, Claude, Gemini, Perplexity, Grok, DeepSeek, Mistral), producing 372 answers. For the small-business segment specifically, no single tool owned the answers: recommendations were fragmented across many names with none dominant. That fragmentation is the opportunity. When no one owns the answer, a business that gets its presence right can move into the shortlist quickly, because there is no entrenched favorite to displace.

A worked example: Ferreira & Sons, a two-van plumbing business

Say Ferreira & Sons is a family plumbing business with two vans, covering a mid-sized city. They have a decent website, a full Google Business Profile with 90 good reviews, and plenty of word-of-mouth work. They have never checked what an assistant says about them. A scan runs the questions their buyers actually ask: "best emergency plumber in [city]," "who can fix a burst pipe fast near me," "good plumber for a bathroom renovation in [city]," and a dozen more, across all seven engines, several times each. Here is what it turns up.

• On the emergency questions, the assistants name three competitors again and again. Ferreira & Sons appears in roughly one answer in ten, and never in the top spot. The winners are not better plumbers. They are simply listed on two directory sites and mentioned in a local "best plumbers in [city]" roundup that the engines keep quoting.
• On the bathroom renovation questions, Ferreira & Sons does not appear at all. Their site has no page about renovation work, so there is nothing for the engine to connect to that question. The competitors who win each have a dedicated renovations page written in plain language.
• Their 90 reviews barely help, because they all sit on a single Google profile the assistants do not quote from directly. On the review sites and directories the engines do read, Ferreira & Sons has an unclaimed stub with no reviews and an old phone number that does not match their website.

The report's first three fixes follow straight from that. One, claim and complete the two directory listings the engines already quote for this city, fix the phone number so it matches everywhere, and ask ten recent happy customers to review them there rather than only on Google. Two, write a plain "bathroom renovation plumbing in [city]" page that answers what it costs, how long it takes, and what is included, so there is finally something quotable for that whole class of question. Three, get into the local "best plumbers" roundup that keeps deciding the emergency answers, by reaching out to the author with their reviews and track record. None of these costs real money. All three attack the exact sources the scan showed were producing the competitors' wins. That is the difference between guessing and measuring: the work list writes itself.

Where to start this week

If you do nothing else, spend an hour doing the manual version. Open ChatGPT, Gemini, and Perplexity, ask each one the three or four questions your best customers would ask before hiring you, and write down who gets named. If you are missing, note which businesses win and go look at where they are present that you are not. That alone will point you at your first fixes. When you want the complete picture across all seven engines with the sources mapped and a prioritized plan in your hands within the hour, that is what the Scan is for.

FAQ