Sitegravity
← All research

Private hospitals in Malaysia: what AI recommends in five languages

Published: November 3, 2026

In a parallel test of private hospitals, the recommended lists were about 63% the same across five languages. The hospital we tracked was named in 44% of answers: 64% to 67% on DeepSeek and Gemini, 22% to 24% on ChatGPT and Perplexity.

What we asked

Three scenarios: a heart check-up in the capital; a knee replacement for an overseas patient choosing between the capital and a northern city; and a patient flying from Indonesia for gallbladder surgery in that northern city. Each scenario was written once per language (English, Simplified Chinese, Traditional Chinese, Malay and Indonesian) and asked three times on each of four AI engines. In total: 180 answers.

Main findings

  • The hospital we tracked was named in 44% of answers (43% to 45% across sampling rounds), and when it was named it was recommended in 98% of cases.
  • By engine: Gemini 67%, DeepSeek 64%, Perplexity 24%, ChatGPT 22%.
  • The same question, very different answers by engine. In the knee-replacement scenario, DeepSeek named the hospital in 14 of 15 answers and Gemini in 15 of 15. ChatGPT and Perplexity named it in 0 of 15.
  • By scenario: 0 of 60 in the capital heart scenario (none of the four hospitals we tracked is in the capital), 48% in the knee scenario, 85% in the scenario about the northern city.
  • By language: Malay 50%, Indonesian 47%, Traditional Chinese 44%, Simplified Chinese 42%, English 39%. Indonesian was not lower: in the scenario about a patient flying from Indonesia, Indonesian-language answers named the hospital in 11 of 12 answers.
  • The recommended lists were about 63% the same across languages (median of 30 language pairs, range 38% to 89%; by scenario 67%, 66% and 60%). Around four in ten recommendations changed with the language.
  • One hospital in the capital was named in 59% of all answers (106 of 180).
  • A channel only one language recommended. In the scenario about the patient from Indonesia, 2 of 12 Indonesian-language answers recommended a medical-intermediary website aimed at Indonesian patients. The answers in the other four languages did not recommend it, although a few cited its web address as a source.

What it means

Patients are told different things in different languages and by different engines. A hospital can be the first name in one engine and absent in another for the same question. For patients travelling from another country, the intermediary websites that AI points to in their language matter as much as the hospital lists.

Method and limits

Each language and scenario has 12 answers; differences of a few points are noise. The overlap counts only hospitals the AI recommended and that appeared in at least 2 answers for that scenario and language. The Indonesian-channel example rests on 2 answers and is a case, not a rule. One test, one date; models change.

Data collected on 3 November 2026. The hospital and its competitors are not named.

Data collected October and November 2026. Findings are anonymised and describe one test on one date. → Method

Want to see this for your brand?

Get a free AI visibility check