Research · Guide
Win more local AI recommendations: 3 moves for service brands
Customers are asking AI for recommendations before they visit your website. Here's how to find out whether you make the list, make sure AI gets your facts right, and test which message moves you up it.
Your next customer may ask an AI assistant for a shortlist before they ever see your website. If you're not on it, your landing page never gets the chance to persuade them.
This guide is for marketers at local-service businesses: clinics, salons, studios, repair shops. By the end, you'll know the three questions to answer before you spend more on SEO or GEO:
- Does AI recommend you when customers describe what they need?
- Does it get your facts right: branch, service, price, availability?
- Which brand message would earn you more recommendations, before you pay an agency or creator to amplify it?
Why start now
AI shopping is arriving in stages, and the first stage is already here.
- Discovery is mainstream. 74% of respondents across 14 Asia Pacific markets use AI tools to discover or learn about products (Visa/YouGov, 14,764 adults, September 2025).
- The traffic is growing fast. AI-driven traffic to retail sites grew 127% year on year in August 2026 (Adobe).
- Agents are heading for the storefront. Visa projects that a quarter of digital storefront interactions will be started by AI agents by 2028 (Visa).
Trust is moving more slowly. In the same Visa survey, only 14% of Singapore respondents were interested in letting AI make purchases for them. It may be a while before customers hand an agent their credit card, but AI recommendations are already shaping what they buy.
A local service marketing example
Picture someone typing: "Find me a Tuina session near Novena after work. Keep it under S$100, and check availability before booking."
Your clinic might be a perfect fit. But can an AI assistant tell which branch offers Tuina, what a first visit costs and whether there's an evening slot? Agents such as Meta's Muse and Grok Bot are being built to compare options and take the next step. Whether they can do that for your business is something you can check.
1. Find out where you lose the shortlist
Ask the questions your customers ask, without your brand name. "Tell me about our clinic" only shows how AI describes a name you handed it. "Compare Tuina options near Novena MRT with first-visit prices" shows whether you'd be found at all.
Add the constraints real customers add: a budget, a time, a single session rather than a package. A clinic that appears for a broad neighbourhood search can vanish the moment someone says "after 7pm".
Ask more than once. AI answers change from one session to the next, so one good screenshot proves little; repeated sampling shows how often you really appear. Then read the pattern:
| What you see | Where to look first |
|---|---|
| You're missing from relevant shortlists | Whether your branch and service details can be found, and what sources back your competitors |
| You appear with the wrong price or branch | Conflicting pages, old listings, expired promotions |
| You disappear when the customer asks for an evening slot | Whether your hours and availability answer that need |
Takeaway: name your biggest gap. Are you not being found, being described wrongly, or being quoted the wrong facts?
2. Give AI evidence worth recommending
Think like the customer. Before booking, they want to know: is it close, is it the service I want, what will I pay, can I go when I need to, and what am I committing to? Answer each question on your branch and service pages.
Compare "Discover premium holistic wellness" with this hypothetical example:
At our Novena branch, a first visit costs S$88, including an assessment, a 45-minute Tuina session and taxes. You can book a single session without a package. Check the booking calendar for available times.
The second version lets a customer, or an AI assistant, match the clinic to the request. Use facts your team can honour, and have a practitioner review any clinical claims.
Then make those facts agree everywhere: your website, Google Business Profile and booking provider. Contradictions give AI reasons to hedge or get you wrong, so assign someone to update prices and hours when operations change.
You don't need special AI tricks. Google says its AI features rely on the usual foundations, such as crawlable pages, key information in text and structured data that matches the page, and that no special AI markup is required. Its spam policies also cover attempts to manipulate AI responses in Google Search, which rules out the shortcuts: near-identical neighbourhood pages, mass-produced articles, keyword stuffing. One substantive page per real branch beats twenty pages with a different MRT station in the heading.
Takeaway: fix one real branch end to end, from service description through to a working booking link.
3. Test the message before you fund marketing
Being found and described correctly isn't the same as being chosen. AI can describe your business accurately and still recommend a competitor, because your brand story gives it too few reasons to pick you for this customer's need.
Say our hypothetical TCM clinic has built its identity around heritage. AI reflects that faithfully: "a traditional TCM clinic in Novena." But the clinic also offers evening appointments, single sessions without a package and a clearly priced first visit, and those barely show up. For someone looking for a session after work, the best reasons to choose this clinic are invisible.
That becomes a budget question. Should the SEO agency write about expertise, the first-visit experience or evening availability? Should the next KOL brief lead with tradition, convenience or no-commitment booking? Each angle means different pages and different briefs, and each costs money to produce and distribute.
| Angle | What it brings forward |
|---|---|
| Fits around your working day | Evening hours, location, live availability |
| Easy to try without a package | Single-session booking, clear terms |
| Know what your first visit involves | Assessment, duration, full price |
Before you pay to amplify any of these, test it in a controlled environment. Show an AI model the same customer requests with your current description, then with the new angle, and compare how often you make the shortlist. A clean result depends on careful design, so consider a research team such as Windtunnel to plan and run the Message Lift test for you.
What comes back is a direction. One angle might lift recommendations for after-work requests and do nothing for the rest. Another might do nothing at all, which is worth knowing before you've paid for it. Either way, you'll know which angle has earned first call on your content, SEO or GEO budget. The test shows what happens once AI has the information; getting it there is still the job of your content and creators.
Takeaway: put the gap, the candidate message and its test result into one brief. Roll the winning angle out small, then check again what AI says about you.
Where to start
Pick one service at one location. Ask AI what your customers ask. Fix the facts it gets wrong. Then test which message gives it a better reason to choose you, before you spend to amplify it.
The clinic and its details are illustrative. Sources reviewed 1 October 2026.