Google Assistant stops working from September 4 on phones, tablets, Wear OS watches, headphones and Android Auto. Gemini takes over that entire install base, and the switch is final: once a device moves across, there is no way back to the old assistant. Google is trading a deterministic system polished over a decade for a probabilistic language model, on tasks where reliability mattered more than conversation.
Key Takeaways
- Google Assistant shuts down from September 4 on mobile, watches, headphones and Android Auto.
- The move to Gemini is irreversible device by device, with a rollout spread over several weeks.
- Interpreter Mode and the daily smart home briefings do not survive the migration.
A hard cutoff date with no way back
The timeline is now public. From September 4, the legacy assistant disappears from Android phones and tablets, Wear OS watches, compatible headphones and cars running Android Auto. Google confirmed the switch in its official note on upgrading the mobile Assistant experience to Gemini.
The rollout does not land overnight. It starts on September 4 and takes several weeks to reach the whole fleet, device by device, with no way for users to pick their own date. One exception holds: cars running Google’s built-in system keep Assistant past the deadline.
One word frames the entire operation: irreversible. Once Gemini lands on a phone, a tablet or a paired accessory, users cannot restore the old assistant on that device. That lock changes the nature of the migration, which is not an update on offer but a replacement being applied.
The deadline arrives a year behind the original plan, which targeted 2025. That slip reads differently next to the model side of the house, where Gemini 3.5 Pro slipped again while Google rebuilt it. The assistant switch was clearly waiting for the conversational layer to hold up under load.
The volume at stake explains the caution. Gemini already closes in on one billion monthly users, and absorbing the Assistant fleet adds an install base that never picked Gemini in the first place. This is audience growth through product retirement, not through stated preference.
What voice control loses in the handover
The migration comes with a feature bill. Interpreter Mode, which handled live translation of a conversation, does not make the trip. Neither do the daily smart home briefings.
The heavier loss sits deeper, in how the new assistant actually works. Gemini is a language model that generates its answer word by word, which adds 380 to 620 milliseconds of cloud latency before it even starts acting on a command. On a conversational query, that delay goes unnoticed. On a “turn off the living room light” shouted from the hallway, it shows.
That gap is the whole difference between the two philosophies. Google Assistant ran on deterministic routing: a recognised phrase triggered one precise action, every single time. Gemini produces a probabilistic answer, looser about phrasing, weaker on the guarantee that the same request executes identically twice.
For anyone with a smart home setup, the trade is concrete. They gain an assistant that can hold a follow-up exchange and reason through a layered request, in line with a Gemini 3.5 Flash that already sees a screen and clicks on their behalf. They lose the certainty that a simple command produces exactly the same effect, at the same speed, every morning.
Product teams that built voice routines on Assistant therefore have a verification job ahead of them. Every home automation scenario, every watch shortcut, every in-car interaction has to be retested after the switch, with no rollback if the result disappoints. That full handover to Gemini extends the redesign started with the new look given to Gemini at the developer conference, this time all the way down to everyday devices.
More articles on Horizon
- Pulitzer Prize 2026: Eight Awardees Disclosed AI Use
- ChatGPT in Congress Takes 88% of AI Spending
- EU AI Act Enforcement Begins on August 2
Google closes its fleet before the next round
Beyond the technical migration, the move settles a strategic question. Google was running two competing assistants inside its own ecosystem, one of them a legacy product nobody was really developing. It keeps one.
The timing lines up. By unifying its fleet before year end, Google puts its consumer assistant on the same software brick as its agents. One model to evolve, one voice entry point to defend against rivals.
On the competitive side, pressure ticks up on Apple. Siri is being rebuilt at the exact moment Google retires its legacy assistant, which hands Cupertino an easy pitch to users who valued reliability on simple commands. The coming months will show whether that pitch travels beyond the circle of the annoyed.
For connected device makers, the problem bites harder. They certified speakers, headphones and watches against a voice layer that is disappearing, and now have to validate their integrations on an assistant with less predictable behaviour, without being able to offer customers the option of staying put.
Then there is pricing, never far away when an install base moves in one go. Google has already shown it will fight on that ground, dropping its AI Plus plan to $4.99. A fleet unified on Gemini is also a mass conversion channel toward paid tiers, once the voice habit has transferred to the new model.
Follow the story on Horizon.
