Google ships new Gemini Live voice features as its AI feature naming draws criticism

What Google announced

Google used a Wednesday announcement to roll out new Gemini Live voice features, framing the update around the idea that people should no longer have to work out in advance which tool their request belongs to. According to TechCrunch, the company's own copy states that users "shouldn't have to guess whether a task requires Spark, a Daily Brief, or a quick inbox search."

As TechCrunch observes, there is an obvious tension in that line: each of those capabilities is a separately branded feature inside the Gemini app, which is exactly what makes guessing necessary in the first place.

Three destinations inside one app

The Gemini app currently splits chat, Spark and Daily Brief into distinct areas, each with its own icon and slot in the navigation. TechCrunch argues this fragments what ought to be a simple consumer experience, and reads it as a sign that Gemini still lacks a defining use case.

Daily Brief, as TechCrunch describes it, works like an AI-generated agenda that delivers "proactive, personalized updates" drawn from Google services such as Gmail and Calendar. The complaint is that it cannot reliably tell urgent, actionable information apart from unrequested nudges — reminders to pick up earlier chatbot research, for instance, or the resurfacing of past Google searches. That last behavior, in TechCrunch's assessment, lands as invasive rather than useful: having once searched for college scholarships or animal rescues does not mean you want an assistant raising the topic again later.

Spark has the opposite problem, in TechCrunch's telling. It is one of the more capable parts of the app — an agent that can take action on your behalf — but Google has given it a standalone identity it does not need. Internal team branding may matter inside the company; a mainstream user should simply type a request and let the system decide whether a task warrants spinning up an agent.

Not just a Google problem

TechCrunch is explicit that Gemini is not alone here: the AI industry as a whole tends to expose its internal architecture to consumers rather than hide it behind a simpler front end. Anthropic's Claude app asks people to choose between "chat" and "Cowork," and until this week, TechCrunch notes, the two modes did not even share a memory of past conversations. OpenAI's ChatGPT likewise separates "Chat" from "Work." The piece's conclusion is that this engineering-minded design makes AI feel unnatural to use — people are being asked to memorize brand names for what are essentially interaction modes running on the same underlying models.

The simpler counterexamples

TechCrunch points to two approaches that sidestep the issue. Apple's low-key Siri strategy may ultimately win people over because it asks nothing of them: Apple simply makes the things people already use — Spotlight search, Photos, the iPhone camera, Siri voice requests — smarter, without introducing a new interface to learn.

At the same time, a crop of text-based assistants — TechCrunch lists Poke, Ollie, Lindy, Orchid, Lucas, Folk, Tomo and Instinct among others — let users simply send a message and have the task handled. Messaging is a clean, well-understood interface that requires no mental effort to work out which product inside a larger app you are supposed to address. As a16z investment partner Justine Moore put it in a post TechCrunch cites: "People don't want to open an app every time they need help – they want a contact they can text like a friend. And the gold standard is iMessage."

Why it matters

The critique goes to how AI products will be judged. If adoption depends on people folding an assistant into everyday routines, then naming and navigation are not cosmetic details — they are the interface. Every sub-brand a user has to learn is friction, and Google's own announcement copy now effectively concedes that guessing between features is a real failure mode.

There is also a trust dimension. Daily Brief reaches into Gmail, Calendar and even search history to push suggestions, and TechCrunch's reaction — describing the resurfacing of old searches as "creepy" — shows how quickly proactive help can read as surveillance when the suggestions miss.

The wider signal for the industry: vendors that hide their architecture and meet users inside interfaces they already understand, whether that is Spotlight and Siri or an ordinary text thread, may hold an advantage over those shipping a portfolio of named modes. Google has the components — a chat surface, an agent, a briefing — but until the app can route a request on its own, the burden of knowing what Spark is still sits with the user.