Google sets September 4 end date for Assistant on Android
Google began emailing users that Assistant stops working on phones, tablets, Wear OS watches, headphones and Android Auto on September 4, with Gemini taking over.
Google has notified users by email that Google Assistant will cease functioning on September 4, 2026, completing a migration to Gemini across most of the company's consumer hardware surfaces.
Scope of the shutdown
The cutoff covers Android phones and tablets, Wear OS watches, Assistant-enabled headphones and Android Auto. The one significant carve-out is vehicles with "Google built-in," where Assistant continues for now. In parallel, Gemini is expanding onto Google TV, Google Home speakers and smart displays — surfaces where Assistant has handled routine tasks for close to a decade.
The date closes out a transition Google originally signalled for 2025 and then postponed. Assistant launched in 2016 and became the default voice layer on well over a billion Android devices; the September deadline retires that layer rather than running the two systems side by side.
Open questions
The substitution is not like-for-like. Assistant was built largely on deterministic intent parsing and a fixed command grammar, which made it predictable for narrow tasks such as setting a timer, toggling a light or starting a call. Gemini is a probabilistic language model, and reporting on the transition has centred on whether it can match Assistant's reliability and latency on exactly those high-frequency, low-complexity operations. Smart-home control is the most exposed category, since a failed light command is more visible to a household than an imperfect summary.
Why it matters
This is one of the largest single-day replacements of a rule-based system with a generative model in consumer computing. It removes the fallback option: from September 4, users on affected devices have no route back to the older, narrower assistant. For Google it consolidates voice, search and agentic actions behind one model family and one data pipeline, which simplifies where new capabilities land. For the wider industry it becomes a live test of whether LLM assistants are yet dependable enough to carry utility workloads at billion-device scale — and the answer will inform how aggressively Apple, Samsung and Amazon retire their own legacy assistant stacks.