Gemini app splits into three: Chat, Spark, and Daily Brief

Chat, Spark, and Daily Brief: three forks inside Gemini

The Gemini app is a big service — Google CEO Sundar Pichai said on August 11 that it had crossed 1 billion monthly active users. Yet inside that single app, ordinary Chat, Spark, and Daily Brief each occupy their own icon and their own slot in the navigation.

Daily Brief is a kind of AI-assistant diary that pulls from Gmail and Calendar to deliver proactive, personalized updates. The trouble is it doesn't distinguish between something urgent and something that's just noise. Critics point out that it can nag you to continue a search you'd started in the chatbot, or resurface something you'd once googled — and it ends up feeling intrusive rather than useful.

Spark is the opposite case. It's an agentic feature that carries out tasks on a user's behalf, and it's widely seen as genuinely useful. The problem is that Google chose to spin it off as a separate brand at all. Ordinary users shouldn't have to think about which part of the app a given request belongs in — but that's exactly what the current structure demands.

Claude and ChatGPT face the same dilemma

This structural problem isn't unique to Google. As the table below shows, Anthropic's Claude and OpenAI's ChatGPT have each carved their products into separate modes too.

Service	Mode split	Notes
Gemini app	Chat / Spark / Daily Brief	Three features, each with its own icon and menu
Claude	Chat / Cowork	Until this week, the two modes didn't even share conversation memory
ChatGPT	Chat / Work	Requires switching modes depending on task type

The Claude app is split into "Chat" and "Cowork," and until this past week, the two modes didn't even share memory of previous conversations. ChatGPT has the same structure, forcing users to toggle between "Chat" and "Work." Regardless of how each company organized its features internally, users end up having to memorize separate brand names for what are, in the end, just different screens built on top of a single model.

Apple's opposite solution

Apple's approach with Siri looks comparatively plain — but that plainness may be exactly what wins over consumers, according to some analysis. iPhone and Apple device users don't have to change a single habit. Spotlight search, the Photos app, the Camera, and voice requests to Siri simply get smarter under the hood; none of it asks users to learn a new interface.

AI assistants that end with a single text

The same logic, some argue, explains why text-based AI services are catching on. Poke, Ollie, Lindy, Orchid, Lucas, Folk, Tomo, and Instinct all work the same way: the user just sends a text, and the assistant handles the rest. In a post from a16z investing partner Justine Moore, she writes that people "want a contact they can text like a friend." Texting is an interface everyone already knows, which means there's no need to figure out which feature to use inside some larger app.

The same branding fight is playing out in Korea

According to domestic reporting, Google has once again rolled out a promotion offering Korean university students a full year of free Gemini paid membership to coincide with the new fall semester, and the marketing battle between ChatGPT and Gemini is heating up in the Korean market as well. While OpenAI strengthens its dedicated environment for teens aged 13 to 17, Google is targeting university and graduate students — and some see both companies turning classrooms into their next battleground. But the fiercer that marketing competition gets, the more likely it is that the same confusion — users not knowing which button to press inside the app — repeats itself in the Korean market too.

Editor's take

What matters here is that this isn't just a Google problem. ChatGPT, Claude, and Gemini are all, in effect, selling not one language model but a collection of separate windows built on top of that model — and all three have designed their products by slapping a distinct name on every single window. It makes perfect sense for an engineering org to split into a Spark team, a Cowork team, a Work team. But it's a different matter entirely to expect users to learn that org chart by heart.

Looking at it across generations makes the pattern clearer. Back in February 2024, when Google folded Bard into Gemini, the pitch was simplicity itself — one chatbot, one name. Two years later, that same single app has spawned three sub-brands again: Chat, Spark, Daily Brief. Rather than getting simpler as features multiplied, it fractured further — the same pattern playing out as ChatGPT split into Chat and Work, and Claude into Chat and Cowork.

For teams in Korea looking to adopt AI tools, there's a practical lesson here: drop the habit of naming every new feature the moment you build it. When rolling out an AI assistant internally, resist the urge to split the interface into a search screen, a summary screen, a drafting screen. Aim instead for something where users type one thing into one box, and whatever capability is needed kicks in automatically behind the scenes. That's ultimately why text-first AI assistants like Poke and Lindy are catching on — the only interface a user ever has to learn is how to send a text.

It wouldn't be surprising if Google reworks the Gemini app's navigation in the coming weeks. This criticism is landing right as Google is pushing its new voice features hard, so the next update may well consolidate the branding or at least simplify the screen structure. On the other hand, if Anthropic and OpenAI move first to erase the line between chat and work modes, Gemini could end up looking like the one playing catch-up.