On June 23, 2025, Google announced that AI Mode — its experimental AI search experience — had arrived in India, the feature’s first market outside the United States. Users opt in through Search Labs, queries are English-only at launch, and both voice input and Lens-based image search are supported. India has more than 870 million internet users and is Google’s largest growth market.
For search and product teams, the interesting part is not just another country on the map. India is one of the highest-penetration markets for voice and visual queries — Google says Lens has more monthly users there than anywhere else. What AI Mode will validate in India is how AI search behaves in a non-Western, multilingual, mobile-first market, which is exactly the data Google needs before any wider global rollout.
Opt-in through Search Labs
AI Mode in India ships as an experiment: users have to join Search Labs and switch it on inside Google Search — and Search Labs is itself a self-selection mechanism, surfacing users willing to tolerate rough early features in exchange for priority access. TechCrunch reported that AI Mode handles complex, multi-part queries and supports follow-up questions to refine results — something ten blue links never did — and that only English queries are supported at launch, with Google not saying when or whether local Indian languages will be added. The timeline so far: AI Mode was tested with US premium subscribers in early 2025, opened to all US users after Google I/O, and India is the first international market. The cadence makes clear Google is treating AI Mode as an ongoing experimental product line rather than a one-time feature launch.
Custom Gemini 2.5 and query fan-out
According to Google’s own announcement, the Indian version runs on a custom version of Gemini 2.5. The core technique is “query fan-out”: the system breaks a question into subtopics, runs many searches simultaneously, and synthesizes deeper web content into an answer. Developers should study this architecture — turning a single query into a set of parallel subtasks, integrated by a model, closely resembles the task-decomposition patterns used in AI agent systems. The system also draws on the Knowledge Graph and shopping data, and falls back to standard web results when confidence in the AI answer is low. Google’s early numbers: AI Mode users issue queries 2 to 3 times the length of traditional searches — longer, more complex questions are precisely the reason this product exists.
Voice and images, tuned for India
The feature choices are clearly aimed at local behavior. Voice queries work through the microphone icon, and image search is integrated with Lens: users can snap or upload a photo and ask complex questions about what is in it — Google’s example is identifying a plant and getting care instructions. Google notes that voice and visual search are especially popular in India, and with Lens’ monthly users higher there than anywhere else, India is the densest laboratory for visual-query data; input patterns validated here are likely to feed back into product design elsewhere. For product teams, the signal is explicit: AI search input methods should follow a market’s input habits, not ship the US version untranslated and unchanged.
Competition and the traffic problem
TechCrunch placed the launch in its competitive frame: AI Mode is Google’s direct answer to chat-style search products like ChatGPT and Perplexity, and India’s scale makes it a critical battleground — a testing environment whose behavioral-data volume other markets cannot replicate. Google’s own figures add context — AI Overviews now serve over 1.5 billion users monthly and drive a 10%-plus increase in related query usage in major markets including the US and India, which frames AI summaries as a search-volume amplifier rather than just a cost center. But reporting has also noted that AI answers are cutting traffic to publishers. Every market AI Mode enters magnifies the same tension: better user experience on one side, a sharper content-ecosystem distribution problem on the other.
What to watch
Three things are worth tracking. First, when Hindi or other Indian languages arrive — multilingual capability is the mainstream requirement in this market. Second, usage data from India, especially the share of voice and Lens queries, and how it reshapes feature priorities in other markets. Third, how search-traffic redistribution plays out in India, which will be the highest-volume natural experiment yet on AI search’s impact on the content ecosystem.
Sources
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.