The three things the AI weighs first
AI assistants are not scoring your business on a single number. They are answering a specific question for a specific person, and they assemble a short list by checking a few things at once. Three factors do most of the work, and they are easier to act on once you can name them.
If you have ever read about how Google ranks the local map results, this will feel familiar, because the AIs lean on a lot of the same public signals. The labels below are a plain way to think about it, not official terms any AI company publishes.
- Relevance: how well your business matches the exact thing the customer asked for, by category and by the specific need in the question
- Prominence: how real and well regarded you look, judged from reviews, mentions on lists, and how often the wider web talks about you
- Proximity and area: whether you are actually in, or serve, the place the customer is asking about
Relevance: do you match the exact question
Relevance is the first filter, and the most literal. If someone asks for an emergency plumber, the AI is looking for businesses that plainly do emergency plumbing, not general handymen who might. If someone asks for a dentist who takes nervous patients, it wants a sign you handle that. The closer the match between the question and what your sources say you do, the more likely you make the cut.
This is where your category and your own words matter. The primary category on your Google Business Profile tells the AI what bucket you are in. The description, your services list, and the language on your website tell it the specifics. A plumber whose profile and site both say emergency and 24-hour will match an emergency query far better than one whose listing just says plumber.
The fix this week is small and concrete. Make sure your primary category is the one a customer would actually pick, add accurate secondary categories, and say the specific services you offer in plain words on both your profile and your site. Our guide "How to get your local business recommended by ChatGPT" walks through getting these fields right.
Prominence: do you look real and well regarded
Plenty of businesses match a given query. Relevance gets you into the pool. Prominence decides who rises out of it. This is the AI asking, in effect, which of these matching businesses do people actually trust and use.
The clearest signal of prominence is reviews: how many you have and how high they rate. A business with a hundred reviews at 4.7 reads as more established than one with six at 4.5, even if both fit the question. On top of reviews, the AI notices whether you turn up on the kind of lists it trusts: local press roundups, established directories, and industry-specific lists. Being named in a real local best-of piece tells the AI you belong in the answer to best, not just to the bare category.
Prominence is the factor you build slowly rather than fix in an afternoon. You cannot buy it, and you should not try, because there is no paid slot inside an AI recommendation. What works is the steady stuff: keep earning genuine reviews, and get listed accurately on the few directories that matter for your trade. Our guide "How to get more Google reviews" covers the habit side of this.
Proximity and area: are you where they asked
Proximity sounds obvious and trips up more businesses than you would think. When a customer asks for the best in a city or near me, the AI is matching against where you say you are and where you say you work. If your service area is unclear, or your address points to a different town, you can be the best match on relevance and prominence and still get dropped because the AI is not sure you cover the customer.
This matters most for two kinds of business. If you are a storefront, your exact address and the right map pin decide which neighborhood searches you show up for. If you travel to customers, like a plumber or a mobile groomer, your stated service area is doing the work the address would do for a shop. An unset or vague service area quietly shrinks your reach.
Set your service area explicitly. If you work out of a location, confirm the address and map pin are exact. If you travel, list the towns or zip codes you cover on both your profile and your site so the AI has something definite to match against.
The tiebreaker: review recency and what reviews say
Here is the part that decides most close calls. Once two or three businesses are roughly equal on relevance, prominence and area, the AI still has to choose which to name first, or which to name at all. The tiebreaker is almost always your reviews, and not just the count. Two things break the tie: how recent your reviews are, and what they actually say.
Recency is a freshness check. Thirty reviews from the last three months read as a business that is open, busy and still doing good work today. Two hundred reviews where the newest is from two years ago read as a business that might have slipped or closed. When the AI is picking between similar options, the one with fresh reviews looks like the safer thing to recommend.
The text is the other half, and it is the lever most owners ignore. AI assistants read the words in your reviews, not just the stars, and they pull out themes. If your reviewers keep writing that you do same-day repairs, that you are good with anxious kids, or that someone on staff speaks Spanish, the AI can match those exact phrases to a customer's exact question, even when your overall rating is no higher than the shop next door. Your reviews are quietly writing the description the AI uses to decide between you and a rival.
- Ask happy customers for a review steadily, so you always have recent ones, rather than collecting a burst once a year
- When you ask, mention the specific thing you did, so people naturally name it in their review and give the AI a theme to match
- Reply to reviews in your own words, since your reply adds text the AI can read and reinforces what you are good at
- Do not fake, buy, or trade reviews; manipulation gets caught and the penalty lands on you, not the rival you were chasing
A mental model you can actually use
Put the four pieces together and you have a simple way to picture every recommendation. The AI starts with the question, keeps the businesses that clearly match it (relevance), narrows to the ones that look real and well regarded (prominence), drops anyone who is not in the right area (proximity), then breaks the remaining tie with recent, specific reviews.
What is useful about this model is that it tells you where you are losing. If the AI never names you for your category at all, that is usually a relevance or area problem: wrong category, vague service area, facts it cannot confirm. If it knows you exist but always names a competitor first, that is usually a prominence or tiebreaker problem: they have more reviews, fresher ones, or review text that matches the question better than yours.
So before you change anything, figure out which step is failing. Fixing your review habit does nothing if your real problem is that your primary category is wrong and you never match the query in the first place. Our guide "Why isn't my business showing up in AI search?" lines up the common causes against these steps.
Why two AIs can disagree on the same question
You will notice that ChatGPT, Gemini and Perplexity do not always name the same businesses, even for the same question in the same town. That is not a glitch. They read mostly the same public sources, but they weigh them differently and read at different moments.
ChatGPT and Perplexity tend to search the live web when you ask, so they react quickly to a fresh review or an updated page, but they also depend on whatever their search happened to surface that minute. Gemini and Google AI Overviews lean harder on Google's own understanding of your business, which makes your Google Business Profile especially decisive there. The factors are the same; the dial settings differ. If you want the full source-by-source breakdown, see "Where do ChatGPT and Gemini get information about local businesses?"
The practical upshot is reassuring. Because the underlying inputs overlap so much, fixing the basics lifts you across all of them at once. You are not gaming four separate machines. You are correcting one shared set of facts and signals that all of them read.
Confirm it by the rate, not by one lucky answer
Once you have worked on the right step, do not assume it landed. Check. Ask the AIs the questions a real customer would ask, in plain language, and see whether you come up and whether they describe you correctly. The catch is that AI answers shift from one run to the next. Ask ChatGPT the same local question three times and you can get three different shortlists, so a single answer tells you almost nothing.
What tells you something is the rate. Run each question several times across ChatGPT, Gemini, Perplexity and Google AI Overviews, and count how often you are named out of all the runs. Being named in eight of ten runs is a real position. Once in ten is noise. Watch that rate move as you fix the relevant step, and you will know which change actually worked. While you are at it, read what the AIs say about you, because that is where you catch wrong hours, wrong prices, or a stale closed flag stated as fact.
Doing this by hand across four AI tools, several runs each, is tedious, which is why most owners skip the step that proves whether their effort paid off. This is the loop LocalFox runs for you. It asks the questions real customers ask, three times each per assistant, and reports how often you are named, what the AIs get right or wrong quoted word for word, and which competitors they recommend instead and why. There is a free check first that gives you a visibility score and your single biggest problem with no account. You can do the whole thing yourself with the steps above; LocalFox just runs the loop and shows you the rate in one place. Our guide "How to check whether your business shows up in ChatGPT" lays out the do-it-yourself version.