Sitegravity
← All research

Attractions and theme parks in Malaysia: what AI recommends in four languages

Published: November 3, 2026

In a parallel test of attractions and theme parks, the recommended lists were about 40% the same across languages. The park we tracked was named in 5% of answers, and only in the scenario about its own city.

What we asked

Three family scenarios in three Malaysian cities: a rainy day with two children and a budget; an indoor play centre or a water park for a four-year-old; and a day out with older parents who should not walk much. Each scenario was written once per language and asked three times on each of four AI engines. In total: 144 answers.

Main findings

  • The park we tracked was named in 5% of answers (2% to 8% across sampling rounds). It appeared only in the scenario about its own city (7 of 48 answers) and in none of the other two (0 of 96).
  • By engine: Perplexity 11%, DeepSeek 8%, ChatGPT 0%, Gemini 0%.
  • By language: Simplified Chinese 11%, English 6%, Traditional Chinese 3%, Malay 0%.
  • The recommended lists were about 40% the same across languages (median of 18 language pairs, range 33% to 64%; by scenario 46%, 38% and 40%). Around six in ten recommendations changed with the language.
  • Attractions are tied to place. In the scenario set in a southern city, one family resort was named in 81% of answers (39 of 48).
  • Most cited websites: online travel and ticketing platforms.

A measurement lesson

AI names the facility, not the resort around it. A water park inside a theme-park resort was named separately from the resort. Once we counted facilities inside a resort as the same brand, the resort's share in that scenario rose from 17% to 81%. Counting only the resort's own name would have understated it badly.

What it means

For attractions, being named depends on the city the customer asks about and on which platforms list your tickets. Check that each facility you operate is described clearly where AI looks.

Method and limits

Each language and scenario has 12 answers; differences of a few points are noise. One test, one date; models change.

Data collected on 3 November 2026. The park 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