How ChatGPT Decides Which Clinic to Recommend in Bangkok

Patients asking ChatGPT to recommend a clinic in Bangkok expect a direct answer — not a list of links. And ChatGPT delivers one, drawing on everything it knows about Bangkok's clinic landscape. But how does it decide who to recommend?

Understanding that decision process is the first step to appearing in it. This explainer covers the three core signals, how each AI platform weights them differently, and what clinic owners can do to influence each one.

Signal 1: Authority — Who Else Mentions Your Clinic?

AI language models are trained on vast amounts of web content. Clinics that appear in that content — mentioned on healthcare directories, reviewed on reputable platforms, cited in articles, featured in local news — carry more "pre-trained authority" than clinics that exist only on their own website.

Think of it as digital word-of-mouth at scale. If your clinic is cited by 20 external sources, it's a much stronger entity signal than a clinic with an identical website but zero external mentions.

This is the slowest signal to build — but it compounds. Every new external citation makes your clinic a more confident recommendation for AI systems.

Signal 2: Content Relevance — Can AI Match You to the Query?

When a patient asks "English-speaking physio for post-surgery rehab in Bangkok," ChatGPT looks for clinic content that explicitly addresses: physiotherapy, post-surgery rehabilitation, English language, and Bangkok. A generic homepage with a Services list matches none of these terms specifically enough to be confidently cited.

Content relevance requires specificity. AI matches content to queries at a semantic level — it understands synonyms and related concepts — but it still needs explicit signals. A page that never mentions "post-surgery rehabilitation" is at a disadvantage for that query, no matter how good the clinic actually is at it.

For a guide on how to structure content for specific clinic types, see our posts on AI optimization for physiotherapy clinics and dental clinics.

Signal 3: Structured Data — Can AI Parse You Reliably?

Schema markup tells AI systems exactly what your content means. Without it, AI reads your page and infers. With it, you're providing a machine-readable fact sheet: clinic name, medical specialty, address, opening hours, languages spoken, and more.

The difference matters most for precise queries. "English-speaking dentist Bangkok open Saturday" is three separate factual claims the AI needs to verify. If that information isn't structured in your schema, the AI either can't confirm it or presents it with lower confidence — reducing the likelihood of a recommendation.

How Perplexity Works Differently from ChatGPT

ChatGPT's recommendations are influenced by its training data — it "knows" things from before its training cutoff, weighted by how often they appear in reputable sources. Perplexity, by contrast, does live web searches and synthesizes results in real time.

This means Perplexity is more sensitive to your current web presence — recent content updates, current schema markup, fresh reviews. ChatGPT is more sensitive to your long-term authority across many sources.

To rank well in both, you need both: a strong historical presence across external sources (for ChatGPT) and well-structured, current content on your own website (for Perplexity).

Google AI Overviews is a third model — it synthesizes from Google's live index and applies additional YMYL (Your Money Your Life) trust standards for healthcare queries. Google AI requires the strongest E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) signals of the three.

What "AI Share of Voice" Means for Your Clinic

In traditional SEO, you measure visibility by ranking position. In AI search, the equivalent metric is share of voice — how often your clinic is cited across relevant queries, across multiple platforms.

A clinic with strong AI share of voice appears consistently: recommended for back pain queries, sports injury queries, and post-surgery rehab queries — across ChatGPT, Perplexity, and Google AI Overviews. The goal isn't to appear once — it's to be the default recommendation for your clinical category in your area.

This is how Locully measures progress. Not just "you appeared in one query" — but a structured audit across 20–30 relevant queries, tracked monthly, showing your share of voice growing over time.

FAQ

Does ChatGPT ever recommend clinics it's never encountered before?

For queries where ChatGPT's training data is thin, it will defer to more current sources — or express uncertainty. Newer clinics can still gain AI visibility faster than established ones if they build strong, well-structured content quickly, because AI weights content quality, not just historical presence.

Can I contact ChatGPT or OpenAI to ask them to recommend my clinic?

No — AI recommendations are generated algorithmically from training data and real-time web content, not from direct submissions. The only way to influence them is to build the signals described above: external citations, quality content, and structured data.

My competitor appears in ChatGPT results. What do they have that I don't?

Typically one or more of: more external citations, better-structured treatment-specific pages, schema markup on their website, or more FAQ content addressing the specific query. Use our AI search audit checklist to systematically identify the gaps.

How often do AI recommendations change?

ChatGPT's training data updates periodically (major model updates every 6–12 months), but Perplexity and Google AI Overviews update in near-real time based on current web content. Maintaining fresh, current content matters for Perplexity and Google AI in particular.