Illustrating the Gemini App

Paper · Source
Human-Centered Design

Source: Google Design · 2026-01-12

How do you build trust in a tool that’s always evolving? Our AI assistant, Gemini, is constantly learning and adapting, which means traditional design methodologies (mostly linear, predictable, and rooted in control) have to be set aside.

Our designers took on the challenge of charting a new digital landscape. How do you bridge conceptual gaps around AI, especially around how Gemini simplifies information and expands ideas? How do you invite users to explore, play, and search, so they can build confidence using an ever-shifting tool? These questions were guided by a core set of goals: to make Gemini feel intuitive, immersive, approachable, aspirational — and, above all, trustworthy.

While Gemini’s sparkle icon may be the most ubiquitous visual element within the Gemini brand, there are many other elements of the product’s illustration system to be explored. In this deep dive, the designers behind the system pull back the curtain to reveal gradients, responsive containers, and intentional motion that seamlessly work together to create a sense of magic, clarity, and familiarity within the Gemini experience.

Navigating uncharted design territory isn’t new. Consider designer Susan Kare, who pioneered the original Macintosh interface. Using simple visual metaphors, she made abstract digital processes tangible and intuitive for new users: a trash can, a paintbrush, a smiling computer face. Her icons weren’t just pixels; they were bridges between human understanding and machine logic. Gemini faces a similar challenge around accessibility, visibility, and alleviating potential concerns. What is Gemini’s equivalent of Kare’s smiling computer face?

The design team landed on gradients, which gently guide users into the new collaborative world with Gemini. It’s an amorphous, adaptable approach, unlike the static nature of past digital assistants (many of us remember Microsoft’s Clippy), yet it inspires a similar sense of discoverability. Gradients might be much more about energy than “objectness,” like Kare’s illustrations (a trash can is a thing, a gradient is a vibe), but they infuse a spirit and directionality into Gemini.

Gradients are central to Gemini’s visual storytelling, acting as context builders that guide users through the product experience. Designed to convey a transfer of energy and directional momentum, they feature sharp, almost opaque leading edges that diffuse at the tail, acting as clear visual pointers to direct user attention toward what’s most important.

To make the system feel alive, our designers wanted to visualize Gemini’s process of active thinking and synthesis, which helps personify the AI assistant rather than rendering it impenetrable.

Concentrated and diffused gradients are used when transcribing voice.

Gemini’s visual language needed to support its own identity while integrating with Google’s existing aesthetic. The design team looked to the company’s familiar logo for inspiration, focusing on the fundamental shape of the circle. This choice was deliberate, as circles tend to convey simplicity, harmony, and comfort. Even Gemini’s own logo is thoughtfully constructed from the negative space of four adjoining circles.

We explored several foundational shapes to represent Gemini’s “thinking state,” drawing heavily from Google’s design heritage. We experimented with the iconic four-color dots — a symbol of the brand’s rounded, optimistic language — and various Material shapes historically tied to voice and Android’s system UI. By leveraging these established elements, we were able to ground the new framework in a familiar history while ultimately evolving them into a more dynamic, fluid expression of AI intelligence.

The iconic four-color dots serve as a reference point for Gemini's rounded, optimistic language.

Material shapes are softened and blurred to take on an ethereal quality.

Many of Gemini’s interface elements, such as buttons and containers, echo the circle through their rounded corners. These subtle visual cues bring continuity with other Google applications. But perhaps the most impactful use of the circle is how it contains and sculpts Gemini’s signature color gradients. Here, shape acts as a dynamic vessel, concentrating energy into a sharp point before allowing it to blossom outwards—a visual metaphor for Gemini’s ability to process and expand ideas.

People expect consistency from digital systems; they want to learn once and apply that knowledge repeatedly. The challenge for the Gemini illustration team has been to cut through that dynamism, creating a new, yet familiar, visual language.

The illustrations are meant to have a “warm, spatial, rounded quality,” says Anna Sera Garcia, design lead. “We’re always considering how to depict our UI in a way that feels optimistic, delightful, playful, yet also sophisticated. There’s something ethereal — that kind of in-between fuzzy space that reflects our nonlinear process for ideation.” The aim was to make sure the illustrations felt intuitive and frictionless, which is crucial for a product like Gemini that’s so new, revolutionary, and always changing.

Movement in Gemini is not merely decorative; it’s an essential guiding element. Each animation has a defined start and end point, creating a sense of directional flow that mirrors user actions. This sense of responsiveness helps users intuitively understand that the system is working with them. Inner activity within the motion conveys thinking, analysis, and intelligence, making Gemini’s processing feel more transparent. Motion allows users to see information coming together, visualizing Gemini’s conversations and listening abilities.

The speed of movement gives a sense of anticipation, then release.

A rippling motion in a radial gradient can be used for voice waves.

Gradients guide a user’s eye to introduce feature icons.

Icons can be animated to indicate new features.

Lines of inquiry this paper opens 2

Research framings built by reading the notes related to this paper — the questions it feeds into.

How do confidence signals and expertise language mislead users' trust calibration? Can mechanistic interpretability reliably explain model reasoning and guide interventions?