Google's new Home Speaker nails hardware, but Gemini assistant falls short

Google has released its first significant smart speaker update in six years with the $99.99 Home Speaker—a device that excels in build quality but struggles with its AI backbone. While the hardware delivers genuine improvements, the Gemini voice assistant powering it still feels incomplete and unreliable compared to Amazon's competing offering.

Hardware that hits the mark#

The Google Home Speaker is a well-executed piece of industrial design. Its compact size—small enough to fit anywhere yet large enough to deliver decent audio—strikes the right balance. The device ships in soft, neutral colors (jade, berry, hazel, and porcelain) that blend seamlessly into most home environments. Touch controls are responsive and intuitive: tapping the top pauses playback or silences the assistant, while side taps adjust volume with subtle visual feedback.

Sonically, the speaker produces clear, punchy sound suitable for music and podcasts. It falls short against the larger Nest Audio it replaces and trails behind the HomePod Mini in bass response, but outperforms Amazon's Echo Dot Max in some scenarios. When paired with another unit, it offers simulated spatial audio and can sync with a Google TV Streamer for synchronized home theater sound—though that feature works only with Streamer content. The speaker also listens well, with three far-field microphones that pick up commands even across rooms or over loud music.

As a smart home hub, the device supports Matter and acts as a Thread border router, letting users add and control compatible devices directly.

Gemini's inconsistencies undermine usefulness#

The software experience tells a different story. Gemini for Home, Google's conversational AI assistant, works across all Google Home speakers but isn't meaningfully faster on the new hardware despite being purpose-built for it. Basic commands like turning on lights sometimes take 10 seconds to execute—significantly slower than Alexa Plus, which handles identical requests in under three seconds.

More troubling are reliability issues. Gemini confidently provides wrong answers: it once played an entirely different song when asked for a specific track, then picked another wrong song when corrected. When asked about voice options, it insisted none existed despite multiple voices being available. On a Nest Hub Max, it fabricated voice names that don't exist. TV controls are particularly flaky, sometimes working, sometimes claiming content isn't available when it is.

These errors stem from a known trade-off in replacing rule-based assistants with large language models—they're more flexible but less predictable. In the context of smart home control, where reliability matters, the mistakes accumulate and erode the core convenience factor.