Gemini for Home finally clicked for me after I deleted half my automations

by · Android Police

Google has spent the first four months of 2026 pushing updates to Google Home and its Gemini voice assistant at a pace consumers have never seen before.

New automation triggers in January, better command targeting in March, and Continued Conversation in late April have all changed what the assistant can handle on its own.

One benefit of these updates is that many of the custom automations that long-time users built to compensate for Google Home's shortcomings are now redundant.

I discovered this firsthand when I audited my own setup. After cutting down my automation list to fewer than 10 items, my home runs better than it did with twice as many automations.

Related

I tried Google Home's biggest update, and came away unimpressed

It's early days, but Google has a lot of work to do

Posts 2
By  Jon Gilbert

Google Home was functional but not flexible

The old assistant understood commands, not intent

Before Gemini, Google Home's voice controls worked very literally. I told it to turn off the lights in a room, and it killed every smart device assigned to that room, including smart plugs running appliances that shouldn't be off.

Also, third-party devices with unusual names were difficult to control by voice, even if they were compatible. That's because the old assistant relied on exact name matching rather than contextual recognition.

Most users learned not to chain two or three commands in one sentence, since it was simply unreliable.

You had to say one thing, wait for the response, trigger the wake word, and say the next thing. For a system marketed as a smart home, it worked more like a voice-activated prompt.

Custom automations naturally developed as a workaround for these issues. They worked, but they often required tweaks whenever my setup changed, and they only solved one problem at a time instead of fixing the core issue.

The 2026 update cycle

Four months of updates that build on each other

The January 2026 update unlocked a new level of automation for Google Home with 20 new starters, conditions, and actions.

For the first time, automations could trigger based on media playback state, appliance status (running, paused, stopped, or error), volume levels, and brightness thresholds.

February brought predefined voice actions and the ability to delete preset routines like Good morning and Bedtime.

The March update landed command isolation, so the assistant targets just lights instead of every device in a room, reduced latency on common commands, and improved voice-trigger reliability.

Continued Conversation launched in late April, letting Gemini retain context between follow-up prompts without repeating the wake word. Asking "What's the weather?" followed by "How about tomorrow?" now works without restating your location.

None of these updates were enough to sway me on their own, and I remained skeptical through the cycle. But when combined over four months, they changed what the assistant could handle without custom workarounds.


As of May 2026, Gemini for Home is on early access.

The basic voice assistant is free on compatible speakers and displays, but Ask Home, Help me create, and Gemini Live require Google Home Premium starting at $10 per month.

The automation editor improvements are available to all users regardless of subscription.


Related

Gemini could finally fix Google for Home's broken automations

Those struggling to be understood by Google Assistant can speak conversationally to Google for Home starting in October

Posts 1
By  Chris Thomas

What I deleted and why

Automations that Gemini now handles natively

After the March command isolation update, I tested the specific situation that had pushed me to build one of my earliest workarounds.

I had a beverage fridge on a smart plug in the same room as several lights. Under the old assistant, telling Google to turn off the lights in that room would shut down the plug. I was running a custom automation to work around this.

After the update, I told Google to turn off the lights, and when I checked the room, the fridge was still running. That's the exact fix I'd been looking for, so I wanted to see which other automations I could delete.

The next one to go was a Good morning routine I had customized to announce the weather, read my calendar, and tune in to the news. Google's preset never worked well for me, so I built it from scratch with my preferred actions.

With predefined voice actions and Gemini's improved accuracy, I get the same result without maintaining a custom routine.

After this, I removed multistep voice commands that I had split into individual automations.

For movie time, I had separate wake words to dim the lights and start the TV. When leaving home, I also needed multiple voice commands to turn off the lights, some appliances, and lock the door.

Now, Gemini handles compound commands in a single request, so I don't have to keep those routines.

Finally, I deleted the stock Bedtime routine. I kept it out mostly out of habit, but with Continued Conversation letting me add follow-ups without the wake word, the routine was not doing anything I could not do faster by just talking.

Where manual automations still win

Some automations don't need AI interference

I kept my sensor-triggered automations, like the motion sensors for turning on hallway lights and the humidity sensor for my fan.

These automations are simple "if this, then that" logic, and they work best when running on autopilot without Gemini's quirks. There's no reason to add voice control when they fire on their own.

The same goes for appliance-triggered alerts. I'd rather have a notification when my washing machine is done as a preset automation than as something I need to remember to ask about.

Subscribe for smart-home fixes and automation audits

Join the newsletter for in-depth smart-home coverage, practical automation audits, and clear checklists for trimming or rebuilding routines—insights on Google Home/Gemini updates and real-world tips to simplify your setup.


Get Updates

By subscribing, you agree to receive newsletter and marketing emails, and accept our Terms of Use and Privacy Policy. You can unsubscribe anytime.

Security automations are too important to leave entirely to AI.

Gemini can handle arming and disarming alarms by voice, but I'm not comfortable relying on a voice command for something with serious consequences if it fails.

I kept the scheduled automation that arms the system at a specific time of the night.

By the end, I have fewer than 10 of these automations left. Some of them were simplified because the new triggers and conditions from the January update gave me better options than what I had originally hacked together.

Related

5 best Google Gemini smart home tips and tricks you should try

Supercharge your home with Gemini

Posts 1
By  Rajesh Pandey

Fewer automations, way better results

The result of Google's 2026 updates is that Gemini for Home now handles the contextual, conversational interaction that the old Google Assistant could not.

And so, users that built layers of complex automations should now consider auditing that list, since some of those workarounds are solving problems that no longer exist.

While I find it fun to script automations, my setup runs more smoothly with a shorter automation list. I kept automations that should naturally be automated, not the ones I built because the assistant could not be trusted.

Google has been working toward this shift since Anish Kattukaran committed to earning back user trust in late 2025, and for my setup at least, it is landing.