How to get your business recommended by AI

Intro

A customer wants what you sell. They don’t scan ten blue links any more. They ask ChatGPT for a good supplier in their area, or they read the AI answer sitting at the top of Google.

Something in that moment decides which businesses get named. Usually two or three. If yours isn’t among them, you never got the chance to compete on price or service, because you weren’t in the conversation.

So getting your business recommended by AI has quietly become its own job, and it’s a different contest to ranking. Ranking is about position on a page. This is about whether a machine trusts your business enough to put its name behind it. Here’s what those systems actually read, and the moves that get you named.

Key takeaways

• AI systems don’t rank local businesses, they verify them. They name the businesses they can confirm from several agreeing sources.
• Your Google Business Profile is the anchor. Google states it helps you appear in AI responses, not just normal search results.
• Directory listings are back. They’re worth only 7% of classic local pack ranking weight, but three of the top five AI visibility factors are citation-based.
• Recent reviews with owner replies matter more than a big historic total. Silence reads as a business that may have stopped trading.
• Ranking well now and being recommended by AI are drifting apart. We audited a client sitting at number one locally whose profile would give an AI engine several reasons to hedge.
• You don’t need special files or markup. Google says so directly.

What happens when someone asks an AI for a local business?

The engine doesn’t look up a ranked list. It gathers what it can find about businesses that match the request, checks whether those details agree with each other, then names the ones it can state as fact. Confidence decides who gets mentioned, not position.

This is the part most guides skip, and it changes what you should work on. A search engine can afford to put you third. An AI system answering in a sentence or two has room for maybe three names, and it will drop any business whose details it can’t confirm rather than risk being wrong.

So the question stops being “how do I rank higher” and becomes “how easy am I to verify”. Those need different work.

Two things get merged here that are worth separating. A Google AI Overview is the AI answer at the top of Google’s own results, above the usual map pack and blue links. A ChatGPT or Perplexity recommendation is a separate product answering from its own picture of the web. Different surfaces, largely the same underlying question about your business.

That split matters more here than in most markets. Google runs 92.12% of South African search as of July 2026, so the Google surfaces carry the volume and the standalone assistants are the growing extra.

Why is your Google Business Profile the centre of this?

Because it’s the cleanest record of your business that these systems can read. It states your name, location, category, hours and services in a structured form, and it’s tied to a review history that’s hard to fake. It’s the reference point other claims get checked against.

Google is explicit about this. Its own guidance on AI features says using Google Business Profile “can help your products and services to be visible in both AI responses and other Google Search results”. That’s the platform stating the link between your profile and AI answers, not an agency’s theory.

There’s a practical reason it carries so much weight. Everything else about your business is a claim you make. Your profile is a claim Google has partially verified, sitting next to reviews from people who don’t work for you.

Google’s local ranking guidance names three factors for classic local results: relevance, distance and prominence. Relevance and prominence both come mostly from the profile and what surrounds it. You can’t move your premises, so that’s where the work goes, and the same material feeds the AI layer.

What signals are these systems checking?

Four things, roughly in this order: whether your details match everywhere, what your reviews say and how recently, whether independent sources mention you, and whether your own pages can be read and understood.

Details that agree. Your business name, address and phone number should be identical across your site, your profile and every directory you appear on. Not similar, identical. A different phone number on an old listing gives an engine two versions of you and no way to know which is right.

Reviews, weighted to recent. Reviews are the trust signal you can’t write yourself. Google confirms that more reviews and positive ratings can help local ranking, and review signals now carry roughly a fifth of local pack weight in Whitespark’s 2026 ranking factors study. Recency does the heavy lifting. A steady trickle with replies reads as a business that’s trading. Two hundred reviews that stop eighteen months ago reads as one that might not be.

Independent mentions. This is the shift most businesses have missed, and it reverses a decade of advice. Directory listings and citations fell to about 7% of classic local pack weight, so plenty of agencies quietly stopped doing the work. In the same 2026 study, three of the top five AI search visibility factors are citation-related, and Whitespark’s own summary puts it plainly: in AI search, mentions are the new link. Being named on an expert roundup or a credible local directory is now one of the strongest ways to be recommended.

Pages a machine can read. Clear service pages, straightforward answers to real questions, and no technical barrier stopping the crawlers. You don’t need anything exotic here, which brings us to the thing you can skip.

Google’s AI guidance states you “don’t need to create new machine readable files, AI text files, markup, or Markdown” to appear in its AI features, and that structured data isn’t required for generative AI search. So if someone is selling you an llms.txt file as the way into AI answers, that isn’t what the platform says. We covered the same ground in do you need an llms.txt file.

What should you actually do, in order?

Start with the profile, because it’s free, it’s yours, and it moves fastest. Then fix your details across the web, then build reviews and mentions, which are slower but harder for a competitor to copy.

1. 01.

   Fill your Google Business Profile completely.

   Correct primary category, secondary categories, services with real descriptions, hours, service area, photos, and a description written the way customers speak. Empty fields are missing facts, and missing facts are what makes an engine hedge.
2. 02.

   Check your own address field character by character.

   This sounds beneath mentioning. It isn’t, as the audit below shows.
3. 03.

   Make your details identical everywhere.

   Your site, your profile, and the South African directories that carry local businesses: Brabys, Yellow Pages, Cylex, Snupit, plus any industry body or chamber you belong to. Fix the old listings you forgot about, because those are usually the ones disagreeing with you.
4. 04.

   Get reviews arriving steadily, and reply to all of them.

   Ask after every completed job. Reply within a day or two, including to the bad ones, because a handled complaint reads better than an ignored one.
5. 05.

   Earn independent mentions.

   Local roundups, supplier and partner pages, industry association listings, “best X in [your town]” lists. Citations are now a top-five AI visibility factor, and almost nobody is doing this work, which is what makes it worth your time.
6. 06.

   Give each service its own page, answering real questions.

   One page per service, with the questions customers actually ask answered in plain language near the top. Dedicated service pages rank among the strongest AI visibility factors in the same study.
7. 07.

   Check the AI crawlers can reach you.

   GPTBot, PerplexityBot and ClaudeBot need to be allowed in robots.txt. Some hosts changed defaults and started blocking them. A blocked crawler means you’re invisible to that engine no matter how good everything else is, and it’s the cheapest problem on this list to fix.

Why do businesses that rank well still get left out?

Because ranking rewards proximity and category match, while being recommended rewards being verifiable. A business can win on the first test and fail the second, and most owners have no idea it’s happening.

We audited the Google Business Profile of a South African home improvement client recently. It ranks at the top of its local results, so by the usual measure it’s winning. The audit still found a spelling error in its own address, an old one-star review sitting unanswered, owner replies that had stopped a while back, and no business description or services list filled in. Scored across the profile, it came out under halfway.

None of that has cost it the top spot. Google’s local ranking still rewards being close and being the right category, and it is both. But each of those gaps is a reason for an AI system to name a competitor instead, or to hedge and list a few options rather than recommend one business.

There’s a second pattern in our own client data. On one retail client’s Search Console, local “near me” searches pull hundreds of impressions a month while sitting on page two, and they convert at under 1% of the clicks they’re shown for, well below the rest of that site. Compare Handles Inc, a hardware e-commerce client with no walk-in counter, whose local-intent searches click through at more than three times its own site-wide rate.

Being seen wasn’t the problem for the first client. Being the obvious, verifiable answer was. That gap is exactly what the AI layer widens.

How long does this take, and what can’t you control?

Profile and listing fixes can show up within a few weeks. Reputation signals like reviews and independent mentions take one to three months to shift what an AI system says, because these engines need to see agreement repeatedly before they’ll rely on it.

Some things stay outside your control, and any agency promising otherwise is overselling. You can’t choose the wording of an AI answer. You can’t guarantee inclusion for a given prompt, in the same way nobody could ever guarantee a number-one ranking. Answers vary between engines and change without notice, and you can’t move your premises closer to someone searching.

What you control is whether you’re easy to verify. That’s the whole game, and it compounds, because the same work also feeds your map pack position and your normal rankings. The wider timeline picture is in how long SEO takes to work.

Does it matter which AI engine you optimise for?

Not much, and that’s the useful part. The inputs overlap heavily, so the work that gets you named by Google’s AI Overviews is largely the work that gets you named by ChatGPT and Perplexity.

The differences are mostly speed and emphasis, and they’re patterns rather than published rules. In what we and others tracking this have seen, Perplexity leans hardest on recency, so fixes tend to surface there first, while ChatGPT tends to favour thorough, well-structured pages. The one part that isn’t guesswork is Google: its AI features run on the same core ranking systems as normal search, which Google states directly, so they follow your organic and local performance rather than moving on their own.

So don’t build a strategy per engine. Build one verifiable business, weight the Google surfaces first given they hold the majority of local search here, and treat the assistants as the growing addition. We unpack the overlap in SEO vs AI search and what AI search optimisation means.

How does red&white approach this?

We start by finding out what the engines currently say about you, then fix the reasons they hedge. That means a profile and listings audit first, because that’s where the cheap wins hide, then reviews and independent mentions, which take longer and hold their value.

In practice: audit the Google Business Profile field by field, map every place your details appear online and fix the ones that disagree, set up a review routine the business can sustain rather than a one-off push, build the service pages that answer real buyer questions, and confirm the AI crawlers aren’t blocked. Reporting runs on enquiries, calls and direction requests, not ranking screenshots.

This is what our AI search service covers, and because the same signals drive local rankings, it builds on rather than replaces our SEO work. If you’re weighing up providers, the questions worth asking are in how to choose an SEO agency.

How do you know if it’s working?

Ask the engines. Open ChatGPT, Perplexity and Google, ask the way a customer would, and write down who gets named. Repeat it monthly. It’s crude, and it’s still the most direct read available.

Then check the numbers that follow: calls and direction requests in your profile insights, impressions on local searches in Search Console, and enquiries that mention finding you through an AI answer. Ask new customers how they found you, because they’ll often tell you plainly.

Watch for the pattern that defines this era, which is impressions holding up while clicks fall. That usually means you’re being read and summarised rather than visited, and the full diagnostic is in what AI Overviews do to SA traffic.

The bottom line

AI systems don’t rank local businesses, they verify them, and they name the ones they can confirm. So the work isn’t chasing a new algorithm. It’s removing every reason a machine might not be sure you are who you say you are.

Complete the profile. Make your details match everywhere. Earn recent reviews and reply to them. Get mentioned by independent sources. Those four moves feed your map pack position, your organic rankings and your AI visibility at the same time, which is why this is worth doing properly rather than as a side project.

The businesses getting recommended aren’t running clever tactics. They’re the ones a machine can describe without guessing. If you want to know what an AI engine currently says when someone asks for a business like yours, that’s where our AI search work starts. Or talk to us and we’ll show you what it sees today.