Gemini's Southeast Asia Numbers Show What Real Localization Looks Like

Nearly 70% of prompts submitted to the Gemini app in Southeast Asia are in a native language, not English — led by Vietnam at 89%, Thailand at 87%, and Indonesia at 84%. That’s the headline figure in Google’s first Gemini Southeast Asia Report, authored by Sapna Chadha (VP, Southeast Asia & South Asia Frontier) and Samuele Saini (Director, Google Ads Solutions & Insights), and it’s a sharper adoption signal than the region’s raw growth number: active users on the Gemini app there have more than doubled over the past year.

The report’s other data points describe a market whose AI usage pattern doesn’t map onto a desktop-first, English-first template. About three in four requests come from mobile devices, over 40% use voice, photo, or video as the input, and roughly 40% of all queries are generative — the report cites 5 billion images produced with Nano Banana and nearly 1 million songs with Lyria 3 in the region over the past year. Gemini’s SEA-HELM benchmark ranking as the top-performing LLM for Southeast Asian languages is presented as the reason the usage pattern looks the way it does, not a coincidence alongside it.

Google is following the usage data with product: Gemini Spark, the agentic feature that manages tasks across Gmail, Docs, and Slides even when a device is locked, is rolling out in local languages this week to Gemini Advanced (Ultra) subscribers after an English-only launch. For companies planning AI rollouts into fast-growing, mobile-first markets, the report is a concrete argument that language-model performance in a user’s own language is a precondition for adoption, not a nice-to-have layered on after an English-first product proves out — the doubled user base followed the local-language performance, not the other way around.