Canlah AI found that 21% of private clinics in Singapore listed on Google Maps do not link to their own website. Among GP clinics, the share rises to 37.7%.
The findings are based on a review of 3,145 private-clinic listings and 1,287 clinic websites in Singapore, along with a separate test of how ChatGPT cited sources when asked about dentists and aesthetic clinics.
According to the Singapore-based AI search agency, missing website links leave many clinics without an owned online source for AI assistants to cite when patients ask for recommendations. In the ChatGPT test, clinics' own websites accounted for 12 of 20 cited links in answers about Singapore dental clinics and 10 of 15 links in answers about aesthetic clinics.
The test also found that ChatGPT named 13 dental clinics and cited the clinic's own website in the same answer for 11 of them. The figures suggest that first-party clinic websites play a strong role in AI-generated local search responses.
Website gaps
The review found that the gap was not evenly spread across clinic types. GP clinics had the highest share of Google Maps listings without their own website, at 37.7%, while dental clinics were at 23.7%.
Canlah AI excluded beauty spas, public institutions and unclassified listings from the review. The remaining sample covered private clinics across Singapore planning areas, with duplicates removed by place ID.
Among clinic websites that loaded, most offered some way to book. Of 1,103 sites, 86.1% had a booking link or form, while 67.4% linked to WhatsApp.
Only 9% had a self-scheduling tool showing open appointment slots. This suggests many clinics still rely on manual messaging or basic contact forms rather than live online scheduling.
Technical issues
The review also highlighted several basic website issues. Around 38.7% of 1,103 sites marked up the homepage as a clinic using schema.org, while 29.4% of 1,222 sites had no meta description.
Dental websites had the highest share of broken pages. Canlah AI found that 7% of 313 dental sites returned an error or broken page, while 3.8% still displayed placeholder "lorem ipsum" text on the homepage.
These findings suggest that some clinics have incomplete sites or poorly maintained web pages. For businesses competing for visibility in both conventional search and AI-generated answers, such issues can affect what machines can read and surface.
Speed results
Mobile performance was mixed. Across 1,204 sites, the median mobile PageSpeed lab score was 59 out of 100, and 29.2% scored below 50.
The picture was stronger in real-user data, where available. Of the 336 sites with Chrome UX Report data, 60.1% were rated fast, indicating that visitor experience can differ from lab-based test scores.
The methodology combined one mobile PageSpeed Insights lab run per homepage with Chrome user data for sites with enough traffic to generate real-world measurements. The agency checked each linked homepage for a booking path, a WhatsApp link, clinic schema.org markup and a meta description.
The findings come as consumers increasingly use AI assistants to search for local services online. These tools often draw on business websites, map listings and other online references when generating recommendations.
In healthcare, where clinics often compete in dense local markets and service categories can be highly specific, a missing or incomplete site may reduce the information available to both search systems and AI tools. The study suggests that digital basics such as working pages, machine-readable clinic details and direct site ownership remain important in that environment.
"More patients now ask an AI assistant which clinic to visit. The assistant can only recommend a clinic it can read about, and one in five clinics on Google Maps gives it no page of their own. The fix isn't glamorous. A clinic needs a homepage that works, pages for its treatments and doctors, and clinic details that machines can read, all within MOH's advertising rules," said Haoyang Pang, founder of Canlah AI.