Google starts removing Google Assistant from Android phones, tablets, Wear OS watches, headphones and Android Auto on 4 September 2026, and replaces it with Gemini. If your product uses Assistant as a voice entry point - an App Action, a voice shortcut, a routine, or anything a user reaches by saying "Ok Google" - that entry point disappears. The rollout takes a few weeks to reach everyone, and once Assistant is gone from a device, users cannot switch back.
What changes on 4 September 2026
The removal covers Android phones and tablets, Wear OS watches, headphones, and Android Auto projected from a phone. Assistant is replaced by Gemini on eligible devices.
Three things are not part of this phase. Cars with Google built-in keep Google Assistant beyond 4 September 2026. Google TV, Google Home speakers and smart displays are not included in the mobile shutdown. And the removal is staged rather than instant, so two people on the same handset model may lose Assistant weeks apart.
The detail that matters commercially is that there is no reversal. This is not a default a user can change back in settings. Once Assistant is removed from a phone, tablet or paired accessory, it is gone.
What this article said in 2016, and what actually happened
This post was first published on 18 October 2016, weeks after Google introduced Google Assistant. Its argument was that Assistant could hold context across several questions in a row, and that this moved voice away from rigid commands towards something closer to conversation. That part was right, and ten years later it is unremarkable - natural language processing good enough to follow a thread is the baseline now rather than the headline.
What the original did not anticipate is how little of the packaging would survive. Google Assistant was itself the successor to Google Now. The demonstration the 2016 article led with was Assistant appearing inside Allo, Google's messaging app - Google shut Allo down on 12 March 2019. And now Assistant, the product that replaced Google Now and outlived Allo, is being replaced by Gemini.
That is three branded surfaces in ten years for one underlying capability. The capability survived every transition and improved at each one. The branded surface a product actually integrates against survived none of them. Keeping those two things apart in your own architecture is the whole lesson of this shutdown.
What to check in your own product this month
The question is not whether your app uses artificial intelligence. It is whether anything in your product, or anything you have told your users, depends on Assistant specifically.
- App Actions, shortcuts and Assistant deep links registered by your Android app.
- Voice paths in a Wear OS companion app, which are affected on the same date.
- Voice interaction in Android Auto that is projected from the phone rather than built into the vehicle.
- Onboarding screens, help centre articles, support macros and marketing pages that tell people to say "Ok Google" to do something. Copy outlives code, and it is the part that gets missed.
- Any device already deployed in the field - a kiosk, a wearable, a fleet handset - where pushing an update is slow.
- Your support team's scripts, so nobody spends September diagnosing the same fault over and over.
If none of that applies, you have nothing to do, and that is the honest answer for most of the Australian products we look at. This is a headline rather than an emergency unless a voice command is the only path to something a user actually needs.
The rule worth taking from this: never let a third party own the only path to a core action
Voice, chat and assistant surfaces belong on top of an action your product already exposes in its own interface and its own API. Build it that way and a shutdown like this costs you a shortcut. Build the assistant as the only way in and it costs you the feature, plus every user who learned to use it. That principle is the practical half of voice interface design, and it applies just as much to a login provider, a maps SDK or a payment sheet.
This is not a theoretical position for us. PixelForce client partnerships run 7+ years on average, which is long enough to still be holding a product when a platform underneath it is discontinued, and long enough to have done it more than once. We have rescued 15+ platforms in the past 3 years: app ratings recovered from 3.8 to 4.6 stars, crash rates typically cut 50 to 70 percent within 2 weeks, and feature development 60 to 80 percent faster after modernisation.
Move With Us is the closest analogue to a migration nobody chose. We rescued the Move With Us platform (200,000+ users): crash rates reduced 50 percent, app performance improved 40 percent, user satisfaction up 30 percent, and zero business interruption during the fix. Zero business interruption is the number that matters when the timeline is imposed on you rather than chosen - the Move With Us case study covers how that work was sequenced.
How much work is this, really
For most products it is small: an audit of everywhere Assistant is referenced, a fallback path through the app's own interface for anything that was voice-only, and a copy pass across onboarding and support material. That is scoping work measured in days rather than a rebuild.
It becomes a real project in two situations. The first is a product where voice was designed as the primary interface, so removing it is a redesign and not a patch. The second is where the Assistant integration is old enough that nobody currently on the team wrote it - that is technical debt rather than a voice problem, and it is usually the more expensive of the two, because the work starts with reading code before it can start with changing it.
Either way it is ordinary legacy system modernisation: find what depends on the discontinued thing, give it a path you control, and remove the dependency. Staying ahead of announcements like this one is exactly what app maintenance is for, and it is far cheaper than reacting three weeks out.
If you are not sure whether this affects you
Most teams do not have a current inventory of which third-party platforms their product depends on, which is why a shutdown announcement turns into a scramble. If you would like a straight answer on whether Google Assistant touches your product, and what else on that list is heading the same way, get in touch and we will look at it with you.