Why Your Google Home Feels Broken: The Gemini 3.5 Pro Delay

The failure usually does not look dramatic at first. You say, “turn off the kitchen lights,” and the speaker hears you, thinks for several seconds, then either acts late or answers as if the request were unclear. A routine that used to start when a phrase was spoken now drifts behind bedtime by minutes. A device that paired cleanly through Matter or Thread suddenly looks unavailable after the Gemini transition. Some automations that worked under Google Assistant no longer trigger until they are recreated. Since July 1, some users have also reported a sharper timeout problem, with commands hanging for up to a minute; Google has acknowledged that delay issue and said a fix is in progress.[1][2][3][4]

That is the useful starting point for the question of Gemini 3.5 Pro’s delay and its smart-home impact. The delayed model is not a magic single cause for every broken Google Home setup. Some complaints began before the July delay story became public, and many of the examples are user-reported rather than independently reproduced across controlled hardware. But the delay does help explain why the repair still feels incomplete: Google has moved more of the home experience toward Gemini, while the current Home-side stack is still struggling with the kind of fast, consistent command execution a house needs.

The Symptoms Are Specific Enough To Separate

The most common complaint is latency on basic voice control. TechRadar described Google Home users slamming Gemini for painful delays, including reports of simple light-switch commands taking around 7 to 10 seconds and routines lagging by as much as 5 minutes.[1] Android Central reported similar 7-to-10-second waits before Google’s March patch, which is long enough to change the emotional meaning of a command. At two seconds, a smart speaker feels busy. At ten seconds, the person in the room starts deciding whether the light is broken, the Wi-Fi is broken, or the platform is broken.[5]

A second class of failure is refusal or misinterpretation. Android Authority reported cases where Gemini on Google Home appeared to argue with or “gaslight” users around ordinary smart-home commands, refusing at first and then completing the same request when pressed.[2] That is different from slow inference. The device may have heard the user and may have access to the right device, but the assistant’s reasoning or routing layer still fails to commit to the obvious action.

A third class is automation breakage. 9to5Google reported in December 2025 that the Gemini for Home upgrade broke some automations, especially voice-keyphrase automations that had worked under Google Assistant. In that case, the documented workaround was not a model setting or a phrasing trick; it was duplicating the broken automation.[3] That matters because it points away from a purely conversational problem. If a copied automation works while the original does not, something in the migration, stored trigger, or execution record has gone stale.

Matter and Thread complaints belong in a separate bucket. Pairing and reliability problems after the Gemini update have shown up in Google Nest Community user reports, but they should be treated as setup-specific until narrowed by device, hub, border router, app version, and firmware. A cloud model delay can worsen the overall user experience, but a bulb failing Thread commissioning is not the same failure as Gemini taking too long to decide what “turn off the kitchen” means.

Then there is the July 1 timeout spike. Android Authority reported that newer Home device delays could produce up-to-1-minute timeouts, and that Google acknowledged the issue on Reddit with a fix “in progress.”[4] That status label is important. It is not a rumor about a future model. It is a current acknowledged platform problem, and users should watch that fix separately from the broader Gemini 3.5 Pro timeline.

The March Patch Helped, But It Did Not Close The Case

Google did ship a meaningful latency improvement before the July delay report. Android Central covered a major Gemini for Home patch in March 2026, with Google Home CPO Anish Kattukaran saying common voice requests became about 30% to 40% faster.[5] Android Police also covered Google addressing Gemini’s largest Home frustrations around that period.[6]

That patch should not be waved away. A 30% to 40% latency reduction on common voice requests is real operational evidence. If a command previously waited at the edge of acceptability, that improvement could move it back into normal household use.

The limit is in the word “common.” A faster path for frequent voice requests does not automatically repair migrated automations, odd phrase triggers, Matter/Thread onboarding, multi-device scenes, or cases where the assistant refuses a command it should understand. The home is not a chatbot tab where a slow answer is merely annoying. It is a command environment. The assistant has to identify the speaker’s intent, map the phrase to the right home graph, choose the correct device or routine, send the command through the right local or cloud path, and return a result quickly enough that the person does not repeat themselves.

That is why partial latency wins can coexist with a broken-feeling Home. A model can get better at the median command while the household still suffers from tail latency, stale automations, edge-case routing, and hardware-specific failures. The person in the hallway does not experience the median. They experience tonight’s command.

Where Gemini 3.5 Pro Enters The Story

Gemini 3.5 Pro became relevant because of timing and promised capability. Google announced Gemini 3.5 Pro at Google I/O on May 19, 2026, with general availability expected in June 2026. Bloomberg then reported on July 16 that the model was months behind schedule because of coding performance misses, a report summarized by Neowin and other outlets.[7]

CNBC reported that Alphabet shares fell 4% after the delay report, wiping out roughly $200 billion in market value.[8] That market reaction is not evidence that your hallway light command failed because of one unreleased model. It is evidence that investors treated the delay as a serious execution signal.

For Home users, the narrower technical point is more useful. A smart-home assistant needs inference speed, coding reliability, and contextual stability. Coding reliability matters because routines and automations are, in practice, little execution plans: if this phrase or device state appears, run these actions in this order against these devices. Contextual stability matters because “turn off the kitchen” is not an abstract language puzzle; it depends on room names, device names, household membership, home structure, and prior platform state.

The caution is that Google has not explicitly said Gemini 3.5 Pro was intended to power Google Home devices. Current Home features have been described around earlier Gemini generations, while Gemini 3.5 Flash is confirmed generally available. So the strongest supported conclusion is not “3.5 Pro was the hidden Home switch.” The stronger conclusion is that the delayed model removes a plausible near-term source of better speed and reliability while the existing Gemini-for-Home stack is already showing stress.

That distinction keeps the timeline honest. User pain predates the Bloomberg report. Some automation issues were documented in late 2025, months before the public 3.5 Pro delay story.[3] The March patch improved some common voice paths before 3.5 Pro missed its expected June window.[5] The July 1 timeout spike has its own acknowledged fix in progress.[4] The delayed model is best read as a reason the deeper repair may take longer, not as the first domino that knocked over every Google Home setup.

The Polls Show Frustration, Not A Precise Failure Rate

Android Authority’s February 2026 reader poll drew 1,530 votes; 32% rated Google Home “1 (Impossibly bad),” while only 3% rated it 5 out of 5.[9] That is an ugly satisfaction signal, but it is not a representative measurement of all Google Home households. Reader polls overrepresent people motivated enough to vote, and smart-home articles tend to attract users who are already trying to diagnose a problem.

Still, dismissing that poll would be lazy. Self-selecting does not mean imaginary. It means the number should be used as a temperature check, not a population estimate. Paired with dated user reports, an acknowledged July delay issue, and Google’s own March latency patch, it supports the basic operational picture: enough users are hitting enough repeatable failures that “just wait for Gemini to get smarter” is not a responsible troubleshooting plan.

Execution Risk Is Visible, But It Is Not The Diagnosis

Bind AI reported that four senior Gemini researchers left Google for Anthropic during the week of June 21 to June 27, 2026.[10] That belongs in the execution-risk column. If a model is already late and senior researchers are leaving near the same window, users are entitled to wonder whether the Home roadmap is competing with bigger internal fires.

But that figure comes from a single source in the material here, and it should not be made to carry more weight than the actual Home symptoms. A user cannot fix a researcher departure. They can test whether an automation is stale, whether a Matter device is failing at pairing, whether the same command works from the app, and whether the July timeout problem persists after Google’s promised fix lands.

What You Can Do While The Deeper Fix Is Missing

Start by classifying the failure instead of rebuilding everything. A slow light command, a broken voice-keyphrase automation, a Matter pairing failure, and a July-era timeout may all feel like “Gemini broke my home,” but they do not have the same first move.

For broken automations, duplicate before you redesign

If a voice-keyphrase automation worked under Google Assistant and stopped after the Gemini upgrade, use the documented workaround first: duplicate the automation.[3] Do not only open the old automation, edit a word, save it, and assume the stored trigger has been refreshed. The useful test is whether a newly created copy with the same logic fires correctly.

If the duplicate works, the old automation is probably stale or corrupted in the migration path. Keep the copy, disable the original, and avoid changing multiple variables at once. If the duplicate does not work, the issue may be broader: phrase handling, device action execution, account permissions, or a current Home-side outage.

For slow commands, compare voice, app, and device control

A voice command that takes 10 seconds tells you the experience is bad, but not where it failed. Run the same action three ways: by voice, inside the Google Home app, and through the device maker’s app if available. If the device maker’s app is instant and Google voice is slow, the problem is likely in Google’s assistant, home graph, or cloud path. If every path is slow, look at the device, hub, Wi-Fi, Thread mesh, or vendor cloud before blaming Gemini alone.

For households with multiple speakers, test from the speaker closest to the room and then from another one. This is not because the speaker should need perfect conditions to turn off a lamp; it should not. It is because the result tells you whether one device is stale, under-updated, or failing to route commands consistently.

For Matter and Thread, verify the chain

Matter and Thread failures deserve a hardware inventory before a platform verdict. Check the exact device model, firmware, Google Home app version, hub or speaker acting as a controller, and whether a Thread border router is actually present and updated. If the same device pairs in another ecosystem but fails in Google Home, that is useful evidence. If it fails everywhere, it is not a Gemini 3.5 Pro story.

This is also where waiting for a model upgrade may be the wrong lever. A language model can help interpret commands and manage context, but commissioning a Matter device depends on protocol support, firmware behavior, radio stability, controllers, and account linking. Treat those as separate until testing proves they are connected.

For July timeouts, wait for the acknowledged fix before doing surgery

If your worst delays began around July 1 and look like commands hanging for up to a minute, put that in the Google-acknowledged bucket. Android Authority reported that Google said a fix was in progress.[4] In that specific case, deleting rooms, unlinking services, or rebuilding every routine may create more cleanup work than benefit.

Do enough testing to make sure you are not dealing with a local outage, then wait for the fix and retest the same commands. The key is using the same command and same device after the patch path changes. Otherwise, you will not know whether Google fixed the timeout or whether you accidentally worked around it.

Rollback is real, but it can be disruptive

Rolling back from Gemini to Google Assistant may be tempting if a household depends on voice commands for lights, locks, or accessibility routines. Treat it as a disruptive option, not a casual toggle. PCWorld covered Google warning of rough edges as Gemini for Home arrived, and the migration path has not behaved like a clean reversible switch for every setup.[11]

On some setups, rollback may require deleting the Home or rebuilding parts of the configuration. That means device assignments, household members, routines, and linked services can become the cost of escape. If you are already planning a rebuild, rollback may be worth considering. If only one automation is broken, duplication is the lower-risk move.

The Most Useful Answer

The Gemini 3.5 Pro delay likely prolongs Google Home’s unreliable state, especially if the missing model was expected internally to improve speed, code-like execution, and contextual handling. But the evidence does not support treating it as the sole root cause of every broken light, late routine, or failed pairing. The symptoms are older, more varied, and more hardware-dependent than that.

The practical answer is narrower and more useful: treat Google Home as a platform with multiple active failure modes. Duplicate broken voice-keyphrase automations. Verify the device, hub, firmware, app, and protocol path before calling a Matter issue global. Watch the July timeout fix separately. Consider rollback only when the household impact justifies the rebuild risk. The deeper Gemini repair has to be shipped and proven in kitchens, bedrooms, and hallways before it earns back the assumption that a basic command will simply run.

References

1. It's almost worthless: Google Home users slam painfully slow Gemini assistant — TechRadar — Apr 2026
2. You might need to argue with Gemini to get it to work on Google Home — Android Authority — Jan 2026
3. Gemini for Home upgrade reportedly breaks automations for some — 9to5Google — Dec 2025
4. Google says a fix is in the works for Home device delays — Android Authority
5. Hop to it: Gemini for Home's assistance gets a lot faster with this major patch — Android Central — Mar 2026
6. Google fixes Gemini's biggest Google Home frustrations — Android Police
7. Google Gemini 3.5 Pro faces delays over coding performance misses — Neowin
8. Alphabet shares fall on report its Gemini 3.5 Pro AI model is delayed — CNBC
9. Android Authority February 2026 reader poll on Google Home ratings — Android Authority — Feb 2026
10. Gemini 3.5 Pro Slips to July — and Four Senior Google Researchers Just Left for Anthropic — Bind AI
11. Google warns of rough edges as Gemini for Home arrives — PCWorld