Google Gemini can generate a full sequential data color scheme — and evaluate it against accessibility criteria — in a single conversation. That’s the core finding from a detailed walkthrough on UX Collective exploring whether GenAI can handle the surprisingly technical job of data colorization.
Sequential color schemes map ordered data values to a gradient of hues or lightness steps. Getting them right means satisfying multiple constraints at once:
- Perceptual uniformity — each step in the palette should look equally spaced to human eyes, not just be mathematically even
- Color deficiency accessibility — the scheme must remain distinguishable under protanopia, deuteranopia, and tritanopia simulations
- Sufficient contrast — adjacent colors need enough delta to read clearly in small chart elements like bars or map regions
The author used Gemini to both propose candidate palettes and critique them against these criteria — essentially turning the model into a design review partner. This follows a pattern from the same author’s earlier work using GenAI for analogous color schemes and a separate experiment where Adobe Firefly was tested for similar tasks, with mixed results around color theory accuracy.
The practical question isn’t whether Gemini can spit out hex codes — any LLM can do that. It’s whether the output holds up under the kind of structured evaluation that data visualization professionals actually run. Tools like Viz Palette already exist for free color assessment, and earlier experiments showed GenAI getting confused by perceptual uniformity when the prompts weren’t precise enough.
NOTE
Sequential schemes are one of three common data color types (alongside diverging and categorical). GenAI tends to perform worst on diverging schemes, where it has to balance two hue ramps around a neutral midpoint.
The takeaway for design system teams: Gemini can accelerate palette exploration, but treat its output as a draft. Run every scheme through a dedicated accessibility checker before shipping it into production charts.