Google’s biggest AI advantage is the twenty-five years of device and location data sitting underneath Android: collected for ads, not for AI, and now positioned to make context-aware answers possible in a way no competitor can match without owning a phone platform.
The real question is different from the one most AI coverage asks: when an assistant sits between a person and every decision they make, who understands that person’s situation well enough to act on it?
The advantage nobody designed on purpose
Search shows what people want, backed by a quarter-century of query data no competitor has matched. Maps shows where they are. Android shows the device state connecting the two: battery level, Wi-Fi network, motion, nearby Bluetooth devices, app activity.
A fantastic reverse-engineering analysis from RESONEO mapped the communication between Android devices and Google’s backend: location, Wi-Fi, Bluetooth, cellular, battery state, motion sensors, app activity, device configuration. None of it is interesting alone. A battery percentage tells you nothing. Aggregated across billions of devices, it starts describing how the physical world moves.
StatCounter puts Android’s global mobile OS share at roughly 73% as of late 2025, with an estimated 3.9 billion active devices. That distribution is difficult for a model provider without an operating system to replicate.
The moat and the lawsuit are the same asset
This scale exists because Google collected more than users realized. It paid $391.5 million in 2022 to settle with 40 states over exactly that. It’s the largest privacy settlement in US history, and the same infrastructure is now positioned to make Gemini more useful than a stateless chatbot. Ask a search engine “best restaurant near me” and it ranks by relevance, distance, reviews. Ask a context-aware assistant and it can factor in where you are right now, how long you have, what you ordered last time, whether you’re driving or walking.
That gap isn’t one OpenAI can buy its way into. Google built a location-and-device graph to sell ads, got fined for how it did that, and ended up with the exact substrate a context-aware assistant needs. OpenAI has no Android equivalent to acquire and no legal path to catching up on this specific axis in the time this race gets decided.
Where the moat actually breaks
Here’s the part most of this narrative skips: owning the data doesn’t guarantee owning the interface.
If people stop opening Search and start asking Gemini, ChatGPT, or Claude to just complete the task, the layer that matters shifts from “who has the most context” to “who the user opens first.” Android context is worthless to Google the moment someone routes their intent through a different assistant on the same phone. I’ve written before about how AI search is already replacing the click, long before it fully replaces the interface.
OpenAI’s bet is the mirror image of Google’s: thinner on location, dense on intent. Millions of people typing out exactly what they’re trying to accomplish, in their own words, every day. Neither side wins on data volume alone. The winner is whoever converts context into a completed action with the fewest steps, and right now neither has closed that gap in public.
This is also where the interface fight stops being abstract for anyone doing search work. If the assistant becomes the interface, ranking the page stops being the whole test. The model has to already “trust” the entity enough to name it without the user ever clicking through.
In recent AI visibility work, we measured AI Overview share of voice against the category leader: our client sat around 14%, the leader above 55%. The client wasn’t losing on citations, it was still getting pulled into the source list.
The interesting failure mode wasn’t discoverability. The model already knew the brand existed. The failure happened at the recommendation layer: another entity held a stronger position in the category’s Citation Core, which made it the safer name to say out loud.
For fifteen years, SEO meant helping crawlers parse a document. That’s necessary now, not sufficient. The next skill is winning trust: consistent information everywhere your business appears, unambiguous entity signals, citations from sources a model already trusts. Ranking a page and being trusted inside an AI system’s understanding of the world are two different jobs. Most sites are still only doing the first one.
The actual next battle
Google spent twenty-five years building the world’s largest index of information. Android added a second layer: continuous signals about devices, places, and movement.
That combination is real leverage, but it’s not permanent, and it’s not automatically transferable to an AI product just because Google owns both. The company that wins won’t be the one with the biggest dataset. It’ll be the one that turns “the user wants something” into “the user got it” with the least friction in between, regardless of whose model is doing the reasoning.
Till next time 👋
Ilias