Your clinic has been offered "AI search optimisation". What happens after a patient's assistant finds you?

On Friday 23 October one of the clinics we work with forwarded us an email from a marketing agency. It offered a package to help the clinic show up when patients ask ChatGPT or Google's AI where to go, and it said ads inside those answers were on their way. The clinic asked us whether they should take it up.

Our answer was that being found is a reasonable thing to want, and that it is only the first step. A patient who asks an assistant about a health screening wants an answer about price and preparation, a time that suits them, and a booking that actually lands in the clinic's diary. That second part is what we build and run for clinics, and it is where almost all of our engineering went in October.

Where AI search stands in Singapore today

Briefly, because the clinic asked. Google shows ads inside its AI Overviews in Singapore, but its own help page says those ads are not shown for sensitive categories, and healthcare is named among them (Google Ads Help). ChatGPT began showing ads in Singapore from 23 September (OpenAI), and OpenAI keeps them away from conversations about a person's own health (ad policies). Google also tells site owners there is no special file or markup needed to appear in its AI answers (Google Search Central). What gets a clinic into an answer is a site that states plainly what it offers, where, when and for how much.

So the ground is moving, and a clinic does not need to buy its way in yet. The more useful question is what happens once a patient, or a patient's assistant, arrives.

What we build for that moment

An assistant that answers from the clinic's own catalogue. On the clinic's website and on WhatsApp, patients get answers drawn from the clinic's real packages, prices and preparation notes, and the assistant books into the clinic's practice-management system. When a price changes, it changes in one place and every answer follows. We wrote about why that matters in an earlier piece.

A page written for other people's assistants. For the clinics we run, we publish a set of pages written for AI assistants to read: every package with its price, preparation and next open times, each with a pre-filled link to the clinic's own booking form. Give ChatGPT or Google's AI that page and it can read the open times and hand the patient a booking link; Claude can also connect to the same booking service directly. The patient submits the form themselves, so nothing is booked behind anyone's back, and the form re-checks the time against the live diary. A short plain-text summary for AI tools, an llms.txt file (llmstxt.org), sits alongside it with the clinics, opening hours and every way to book. This is early: today an assistant uses these pages when someone hands it the link, and putting a "chat with AI" button in front of patients on the clinic's own site is the next step.

The work after the booking. This is where October went. A confirmation sent on WhatsApp to the person who made the booking. A preparation page linked from the confirmation email. One reminder before an unfinished booking link lapses. Patients switching a consultation between an in-person visit and a call without phoning in. And since late October, voice notes on WhatsApp: a patient can speak their question and the assistant reads it and replies in text.

Staff who stay in charge. The clinic's team sees every conversation in one console. When a person takes a conversation over, the assistant steps back, and staff are alerted each time the patient replies. Bookings that need a human decision are pinned at the top of the list until someone deals with them.

Someone looking after it every week. Each week we review what the assistant got wrong, correct it, and add the case to a set of tests that every later change must pass. Nothing is deployed without review. We call this keeping it running, and it is most of what a clinic is paying for after go-live.

How to tell what you are buying

Whoever you talk to, it is worth asking three plain questions. When a patient's assistant finds us, what does it say about our prices, and where does it get them from? Can the patient book from there, and does that booking reach our diary without someone re-typing it? Who fixes it when it gets something wrong, and how quickly?

A package that improves how often you are mentioned can sit alongside the work above. What we offer is the part where the conversation ends in a booking, built into the systems your team already uses and kept right after go-live.

If you've had one of these emails and want to know what would make a difference for your clinic, talk to us.