Search behaviour by source market: what patients type, and in which language
International patient search behaviour varies by source market, language and stage of decision, and keyword tools capture only a fraction of it. Read your own search console queries by country and the first messages families send, map them against the four questions every family asks, and build content market by market with someone rostered to answer in that language.
A family in Dhaka, Lagos or Tashkent does not begin with your hospital. They begin with a question typed into a phone, usually late at night, usually after a local doctor has said some version of “this may need to be done abroad”. What they type into that box, in which language, and what they do with the results decides whether your hospital is ever in the consideration set at all.
Most Indian hospitals have no idea what that query looks like. We put real effort into our own catchment. We know how people in our city search for a cardiologist, how they phrase a chest pain question, which neighbourhood names matter. For international demand the same team falls back on a handful of English phrases and a guess. That is why international patient search behaviour is worth studying as its own subject rather than as an extension of domestic keyword work.
It is not one behaviour either. It changes by source market, by the language a family is comfortable reading, by the platform they trust, by whether the person searching is the patient or a son working in another country, and by how far along the decision already is. A query typed by a daughter who has just been told her mother needs a transplant looks nothing like a query typed by a man comparing two cities after his sponsor has already approved travel.
What follows is how I read that behaviour without pretending to precision we do not have, and how I turn it into pages, replies and campaigns that a family abroad can actually use.
What international patient search behaviour actually looks like
Strip away the tools and there are a small number of things a family abroad types. They search the condition or the procedure with the word India attached. They search a city, because someone has told them a city name rather than a hospital name. They search for the best hospital for a specific procedure, which is a phrase we dislike and they use constantly. They search cost and price, in their own currency and in ours. They search the practical journey: visa, flights, how long to stay, where the family sleeps. And once a name has been given to them, by a doctor, a relative or a facilitator, they search that name to see whether anything alarming comes up.
The order matters. Early queries are broad and conditional: is this even treatable, and is going abroad a real option. Middle queries compare, and this is where a country, a city and then two or three hospitals get shortlisted. Late queries verify: the doctor’s name, the accreditation, the reviews, the address. Each stage needs a different page and a different kind of proof, which is the whole argument for building proper country landing pages rather than one international page for the world.
There is also a large category we never see in a keyword tool: the query typed into a messaging app rather than a search engine. Someone forwards a hospital’s number into a community group and asks whether anyone has been there. That conversation is search, in every way that matters, and it is invisible to us.
The language question, and why transliteration breaks your plan
English is the working language of medical value travel out of many markets, but it is rarely the language of the first, frightened search. A family may read English well enough for a discharge summary and still type a question in Bangla, Nepali, Arabic, French, Portuguese, Russian, Amharic or Swahili when they are anxious.
Then there is the middle case, which is the one most teams miss entirely. A lot of searching happens in the local language written in Latin script, because that is what the phone keyboard defaults to and what the person is used to typing. The same condition can be typed in a native script, in transliterated Latin characters, in English, or in a mixture of English clinical words with local-language connectors. Four versions of one intent, and your keyword research probably captured one.
The practical answer is not to translate everything. It is to decide, market by market, which languages deserve a real page and a real reply, and which only need a person on the other end who can speak them. A language you cannot answer in is a language you should not advertise in. That is also the argument for building the desk before the campaign, since a multilingual international patient desk is what makes any of this honest.
Why your keyword tools will mislead you here
Keyword volume data for these markets is thin, often stale, and frequently zero for phrases that generate real enquiries every week. Volumes are small in absolute terms even when the commercial value per case is high, so the tool rounds them away. Country targeting inside the tools is crude. Local-language and transliterated variants are usually missing.
Worse, the tools tell you nothing about where the search happened. In several source markets the first stop is a video platform, not a web search. In others it is a social feed, a diaspora group, or a question asked directly of an AI assistant that summarises three hospitals and names none of them. Treating the search engine as the whole of demand will give you a tidy plan for a small slice of reality.
The second trap is assuming a query means the same thing everywhere. The word private carries one meaning in a market where public hospitals are the default and a different one where almost all care is paid for. The word surgery is used loosely in some markets for any hospital procedure. If you read the query without knowing how the word is used locally, you will write a page that answers a question nobody asked.
So I use keyword tools for shape, never for size. They are useful for telling me that people attach the word cost to one procedure and the word success to another. They are useless for telling me how many families in a given city are considering travel this quarter.
The data you already own and rarely read
The better source is your own footprint. Search console data, filtered by country, will show you actual queries that reached your site from each source market, including ones you would never have guessed. It is a small sample. It is also real. Read it by country, not in aggregate, or the domestic volume drowns everything.
Then read the enquiries themselves. The first message a family sends on WhatsApp is the purest statement of intent you will ever get, in their own words, in their own language, with their own spelling. Export a quarter of first messages, sort them by source country, and read them. I have never done this exercise without finding at least one recurring question that our website answered nowhere. Doing this properly means the desk has to capture source country at first touch, which is one of the unglamorous reasons an international desk needs to run on your CRM rather than on a coordinator’s phone.
Add the call recordings, the questions the interpreters get asked repeatedly, the queries that arrive through the site search box, and the comments under your videos. None of this is a dataset in the statistical sense. All of it is evidence. Read together with arrivals by market, it is a far better guide than any tool, which is the same argument for reading international demand from data rather than anecdotes.
Four questions sitting under every query
Whatever the phrasing or the language, almost every international query is one of four questions wearing different clothes. Can this condition be treated, and by whom. What will it cost me, roughly, and what could make that number move. How do I physically get there and stay there. And who will look after us when we land in a country where we know nobody.
If you map your content against those four questions per market, you will find the gaps quickly. Most hospital sites answer the first reasonably well, answer the second evasively, answer the third with a paragraph written years ago, and do not answer the fourth at all. The fourth is the one families talk about with each other.
The fourth question is worth sitting with. A family is about to fly to a country where they do not know a single person, with a sick relative, limited money and no idea who will meet them. Everything they read about your hospital is being assessed for one thing: whether a human being there will take responsibility for them. Pages that describe facilities answer nothing. Pages that name the desk, show the people, explain who calls whom and what happens at the airport answer a great deal.
The cost question deserves its own discipline. You cannot publish prices and you should not pretend otherwise, but you can explain how an estimate is produced, what is inside it, what sits outside it, who to send reports to, and how quickly a reply comes. A family that understands the process trusts you more than a family given a number they cannot verify.
Turning queries into a market content map
I build one simple grid per priority market. Down the side: the specialties that actually travel from that market, taken from your own arrival history rather than ambition. Across the top: the four questions above. In each cell, the asset that answers it and the language it exists in. Empty cells are the plan.
Two rules keep it sane. First, one market at a time until the pattern is proven. Second, every asset must have an owner who can answer in that language when the enquiry comes, otherwise you are generating demand you will handle badly. A beautifully translated page that leads to a reply in English forty hours later is worse than no page.
The same grid tells you what to make for video and social, since much of the early searching in these markets is happening there rather than on your site. It also tells you which questions your WhatsApp templates need to answer in the first reply, because the first reply to an international enquiry is doing the work a landing page does at home.
Where AI assistants change the picture
A growing share of the early, broad questions now go to an AI assistant instead of a search box, and the answer a family gets is a summary rather than a list of links. That summary is assembled from whatever the assistant can find and trust about hospitals in India: your own pages, directories, news, accreditation listings, reviews, doctor profiles. If your international information is thin, inconsistent or contradictory across the web, you simply will not appear in it.
This does not need a separate strategy so much as a higher standard on the same things: clear, factual, well structured pages, consistent names and addresses, doctor credentials that match everywhere they appear, and answers written as answers. The work behind getting a hospital cited by AI search is the same work that makes a country page genuinely useful to a human being. You can check where you currently stand with an AI search visibility audit before you assume the worst.
What I would do in the first month
Pick your two largest source markets by arrivals, not by aspiration. For each of them, do four things.
- Pull a quarter of search queries by country from search console, and a quarter of first enquiry messages from the desk, and read them yourself rather than delegating the reading.
- Write down the top questions in the family’s own words, including the transliterated and mixed-language versions, and keep the original spelling.
- Map those questions against what exists on your site today, in which language, and mark the gaps honestly.
- Fix the single largest gap first, and make sure a person who speaks that language is rostered to answer whatever it produces.
That is a month of work for one person, not a project. It will tell you more about international patient search behaviour in your markets than any agency deck, because it is built from what your actual families typed rather than from what a tool estimated. Do it again every quarter. The questions move, the languages move, and the markets you thought you understood will surprise you.
Questions people ask
It is the pattern of how families outside India look for treatment here: what they type, in which language and script, on which platform, and at which stage of their decision. It covers broad early questions about whether travel is even an option, comparison questions about cities and hospitals, and late verification searches on a doctor or hospital name. It differs sharply from one source market to another.
Domestic work has reliable volume data, one or two languages and a single decision maker nearby. International demand has thin data, several languages plus transliterated variants, a decision often made by a relative in a third country, and heavy use of messaging groups and video rather than search alone. The methods carry over. The assumptions do not.
Start with the markets where patients already arrive, using your own admission records rather than ambition. Two markets studied properly will teach you more than eight studied shallowly. Once the pattern holds in those two, extend to the next market that shares a language or a referral route with one you already understand, because the content and the staffing effort carry across.
No, and trying usually produces poor translation and stale pages. Decide which languages get a real page set and a real reply capability. A language you cannot answer enquiries in should not be a language you advertise in. Start with the pages that answer the four questions families actually ask, have them reviewed by a native speaker, and keep them current.
A clearer view of where international demand actually comes from and what it costs to serve, rather than a channel budget defended by anecdote. Reading queries and first messages by market shows which specialties genuinely travel and which are aspiration. That makes the case for spending in one market and stopping in another, and it reduces the effort wasted on enquiries the hospital cannot convert.
Use them for shape, not for size. They are reasonable at showing which words attach to which procedure, such as cost, waiting time or success. They are unreliable on volume in smaller markets, weak on local-language and transliterated queries, and blind to searching that happens inside messaging apps, video platforms and AI assistants. Your own enquiry data is the better source.
Digital or growth should own the analysis, the international desk should own the language and reply side, and both should present the same picture to leadership. It fails when marketing writes pages nobody at the desk can support, or when the desk builds knowledge that never reaches the website. One shared quarterly review of queries and enquiries keeps them aligned.
Treat them as first-class variants. Collect the actual spellings from enquiry messages rather than inventing them, and use them in your research and in the way you phrase page headings and answers. Do not try to stuff them into copy artificially. The main value is recognising the intent behind them and making sure the answering content exists in a language the family reads.
Country-level reporting in analytics and search console, source country captured on every enquiry at first touch, file upload that works on a weak connection, and a site structure that can hold language versions without duplicate content problems. None of it is exotic. It simply has to be specified before the marketing team starts producing pages in several languages.
Partly. Domestic travel from smaller cities shares the pattern of a family choosing a city before a hospital, and of a relative making the enquiry. The language mix and the paperwork are different, and there is no visa or currency question. The discipline of reading real queries and real first messages is identical and worth running for both.
Expect a quarter before content and reply changes move anything measurable, and longer for pages to earn search visibility in a new language. The reading exercise itself pays back immediately, because it usually exposes a question the desk has been answering badly for years. Treat it as a standing quarterly habit rather than a project with an end date.
Once set up, a few days per quarter per market. One person pulls the query and enquiry exports, reads them, updates the gap map and briefs whoever owns content and the desk. The heavy effort is the first pass, where you are building the grid and discovering how little of it your current site covers.
Anything you cannot stand behind across a border: comparative claims about other countries, implied outcomes, prices presented as fixed, and eligibility or visa rules stated as current fact. Rules change and a family may act on what you wrote. Describe the process, say clearly that current requirements must be confirmed with the relevant authority and your international desk, and keep the page dated and owned.
Track enquiries and arrivals by market alongside the content you shipped, and watch the quality of first messages. When families start arriving with reports already attached, asking about the estimate process rather than asking whether you treat the condition, your content is doing its job. Rising enquiry volume with no change in arrivals usually means the pages promise something the desk cannot deliver.
