Smart speakers that talk, but fail to act
The latest iteration of smart home technology, Google's new Home Speaker equipped with Gemini AI, was designed to improve conversational interactions and simplify smart home control. However, during testing, it demonstrated a series of significant flaws. The device ignored requests to play songs, changed the subject in conversations, mistakenly identified the day as Tuesday when it was actually Wednesday, and failed to turn off smart lights, citing the error message: “unable to control or interact with your home’s lighting or other physical devices.” The speaker includes a six-month trial of Google Home Premium Standard, a subscription plan that costs $10 per month and introduces Gemini AI features. Despite the promise of smarter and more intuitive control, the device failed to meet basic expectations.
Similar issues are not unique to Google’s offering. Amazon’s Alexa+ smart speaker, enhanced with generative AI, faces the same challenges. Both devices were marketed as offering more natural and personalized interactions, but users are now experiencing reduced performance on basic functions such as controlling lights or playing music. Companies are asking consumers to pay extra for AI features that are not yet fully developed or reliable. This trend highlights a growing disconnect between what is promised and what is delivered in the smart home space.
Older smart speakers worked better
Before the introduction of generative AI, smart speakers from both Amazon and Google provided a more dependable user experience. For example, the Amazon Echo Show 11 could handle a variety of tasks without issue. It played music, set cooking timers, and offered help with recipe conversions. The Google Mini, now out of production, was similarly effective for smart home control. Users could adjust the brightness of their living room lights or ask trivia questions like “Who plays pastry chef Marcus on The Bear?” and receive accurate answers—Lionel Boyce, in this case. These devices were not just functional; they were reliable and easy to use.
The strength of these earlier models lay in their simplicity. Users did not need to learn complex commands to operate them. Everything worked as expected, and the design focused on efficiency. Over time, however, companies began integrating more advanced AI features into their smart speakers, which started to shift the balance from usability to complexity. What had once been a straightforward and helpful tool became more erratic and less dependable. These changes marked the beginning of a noticeable decline in user satisfaction.
AI features come at a cost
In February 2025, Amazon rolled out Alexa+ as part of a new subscription model. This update offered more natural language commands and additional personalized features such as AI-generated music and voice-based smart home controls. Users could now make restaurant reservations, order food for delivery, request an Uber, or even create unique AI-composed playlists and stories. However, for those who own an Echo device but are not Prime members, there is a $20-per-month subscription fee required to access the enhanced features.
Google followed a similar strategy by introducing Gemini AI as part of its Google Home Premium plan. For $10 a month, users could access Gemini Live, which allows for more natural voice conversations and smart-home control using conversational language. Additionally, there is a $20-per-month Google Home Premium Advanced plan that includes voice-enabled video searches from Nest cameras. Without a subscription, users can still ask basic questions and control smart devices, set timers, and perform similar tasks, as in the past. However, the AI-enhanced plans are marketed as the way forward, even though many users find them to be problematic.
Despite the promises made by both companies, the reality of using these AI-integrated smart speakers often falls short of expectations. While the devices may sound more human and engaging in conversation, they struggle with executing basic functions. Even something as straightforward as reading a calendar can fail. In everyday use, the trade-off appears unjustified. Instead of making control easier, these new models complicate the user experience and demand subscriptions for features that were once included at no extra cost. As users continue to adopt these devices, the challenge remains in ensuring that the AI advancements truly improve usability, not hinder it.