AI Visibility for Local Businesses: Why Your Website Is Not the Problem
If an assistant recommends three restaurants in your city and yours is not among them, the cause is almost never your website. For local businesses, these systems read your listings, your reviews and what local publications say about you — in roughly that order. Your site comes fourth.
This is genuinely good news, because the first three are cheaper and faster to fix than a redesign.
Two examples run through this piece: a neighbourhood pizzeria, and a double glazing company that installs windows and doors. They are useful together because they represent two very different kinds of local decision — one impulsive and reputation-driven, one considered and quote-driven — and seeing both shows that the underlying mechanics are the same even when the buying journey looks nothing alike.
What actually decides a local recommendation
Your Google Business Profile
Gemini inherits Google's infrastructure, including business listings. For a physical business, it often reflects the state of your profile more accurately than the state of your website.
The fields that matter are unglamorous: category, hours, address, phone, photos, attributes. A restaurant with no cuisine type set is asking a system to guess what kind of food it serves, and systems do not guess in your favour when a competitor stated it plainly.
The pizzeria example: a profile listed under the generic category "Restaurant" instead of "Pizza restaurant" or "Neapolitan restaurant" is invisible to the specific question "best Neapolitan pizza near me," even if every dish on the menu qualifies. Photos matter the same way — a gallery of five identical exterior shots tells a system nothing; photos of the actual wood-fired oven and named dishes give it something to describe.
The double glazing example is different but has the same failure mode. A profile categorised as "General contractor" instead of "Window installer" or "Double glazing installation service" means the business simply does not surface for the question it most wants to answer. Service area is often left blank too, which matters more for a business that drives to the customer than one the customer drives to — an incomplete service area can quietly exclude entire towns from ever seeing the listing at all. Attributes like "free quotes" or "certified installer" go unset for the same reason profiles go half-finished everywhere: filled in on day one, never revisited.
Half-complete profiles are extremely common, because the business filled in the essentials on day one and never returned.
Reviews, and what they say
Review count matters less than most owners assume. What the reviews mention matters more.
If forty reviews all describe your pizza as the best Neapolitan in the city, a model has forty pieces of evidence connecting your name to that specific phrase. If forty reviews say "great place, lovely staff", it has forty pieces of evidence connecting you to nothing in particular.
For the double glazing company, the equivalent gap is a review that says "very happy, would recommend" versus one that says "installed our bay window in a single day, no mess left behind, and the final price matched the quote exactly." The second sentence answers three separate concerns a buyer actually has — speed, tidiness, and pricing honesty — in one line a model can lift directly. The first answers none of them.
You cannot script reviews and should not try. You can ask in a way that prompts specifics — asking what someone ordered produces a more useful review than asking how their evening was, and for a trade business, asking whether the final price matched the quote produces something a generic "great service, thanks" never will.
Local publications and lists
The "best pizza in the city" articles published by local magazines, city guides and food blogs are exactly what a model retrieves when asked that question. Being on those lists is the single strongest signal available to a local business, and it is entirely off your own website.
Double glazing has a direct equivalent, just in a different set of publications: trade directories like Checkatrade, TrustATrader and MyBuilder, plus regional "best window companies in [town]" roundups on home-improvement blogs. A model asked "who installs double glazing near me, can you get me a few options" is drawing on exactly these directories the same way it draws on food blogs for the pizzeria question.
Trade certifications function the way review-platform badges do for software: a FENSA certification or a "Which? Trusted Trader" badge is a structured, third-party-verified signal that gets referenced in exactly the kind of comparison content a model retrieves — the local-services version of a G2 Leader badge.
Perplexity makes this visible in seconds: ask it your category question, read the sources it cites, and you have a list of publications to approach.
Then, your website
It still matters, mostly for consistency. If your site says one thing about your hours or your offering and your listing says another, you have introduced doubt where there was none.
Two different buying journeys, same underlying mechanics
A pizza decision is usually made in minutes, mostly on reputation and immediacy — is it good, is it open now, is it close. A double glazing decision unfolds over weeks and usually involves getting several quotes before committing, which means the buyer is actively comparing named options rather than picking the first reasonable one.
Despite that difference, the same three layers decide both: a complete, correctly categorised listing; reviews that state something specific instead of something generic; and third-party sources — food blogs for one, trade directories for the other — that a model treats as more neutral than the business's own claims about itself.
The check that takes ten minutes
- Ask Gemini, ChatGPT and Perplexity for the best [your category] in [your city] — for the pizzeria that is "best Neapolitan pizza in [city]," for the glazing company it is "best double glazing installers near [town]" or "who should I get quotes from for new windows in [town]."
- Note whether you appear, and which businesses do.
- On Perplexity, read the cited sources. Those are the pages deciding the answer.
- Open your Google Business Profile and count the empty fields.
- Read your last ten reviews and note whether they mention anything specific.
Most local businesses find the same pattern: profile incomplete, reviews generic, absent from every cited list. All three are fixable without touching a line of code.
How much each channel actually matters
A rough guide to where to spend the first hours, in order.
The one advantage local businesses have
Competition is finite. A SaaS company competes with everyone on the internet. A bakery competes with the bakeries in its city — usually a few dozen, most of whom have done none of this. A double glazing company typically competes with a dozen or two other local installers, most of whom have an equally half-finished Google profile.
That makes local one of the few categories where a small, deliberate effort moves the answer within weeks rather than quarters. Completing a profile and getting onto two local lists is a realistic month of work with a visible result.
Common mistakes
Listing under a category that is technically true but too broad. "Restaurant" instead of "Pizza restaurant," "General contractor" instead of "Window installer" — both are accurate and both are invisible to the specific question a buyer actually asks.
Asking for reviews with no prompt at all. A bare "please leave us a review" produces bare, generic reviews. A specific prompt produces a specific, quotable one.
Ignoring trade directories because they feel like lead-gen sites, not marketing. For a considered local purchase, a directory profile does the same job a review platform does for software — it is often the first thing a model reads.
Letting the website and the listing drift apart. Different hours, a different service area, or a different phone number on each is a small inconsistency that reads as a bigger one to a system trying to state a single fact confidently.
What not to bother with
Schema markup on a small local site is worth doing but rarely decisive. LocalBusiness schema is a nice signal; a complete Google Business Profile is a much stronger one.
A blog nobody reads. Publishing weekly posts about your industry does very little for local recommendations. The systems answering "best pizza near me" are not reading your blog.
Chasing review volume. Going from forty reviews to eighty changes less than going from forty generic reviews to forty specific ones.
The uncomfortable part
Some of what decides these answers is outside your control. If a local magazine published a "best pizza" list two years ago and did not include you, that page keeps being retrieved. The same is true for a regional "top window companies" roundup that ranked the double glazing company poorly in 2022 and was never updated — the article does not know the business fixed the problem it was criticised for. You cannot edit either page.
What you can do is make sure the next list includes you — which means the work is partly PR, an activity most small businesses have never considered part of being findable. That is genuinely new, and it is where the gap between the businesses that appear and the ones that do not is widening fastest.
