Atlas · DS251 · Future / AI: Dynamic Composition

Multimodal Generative UI

Combines on-demand interface with voice, charts, forms and documents.

Status in 2026: Emerging. Appearing in practice, not yet settled.
Trend 2027 to 2030: Potential (an editorial projection, not a forecast).
The term: Known technical term or standard.

Definition

Combines on-demand interface with voice, charts, forms and documents.

Traits

chat
charts
actions

Typical settings: Analysis, Assistance, Support.

When to use it, and when not

Consider it when the task and motivation match: people asking varied questions. Typical settings: analysis, assistance, support.

Be careful when the risk below would hurt the main task: confusing mix of channels and data.

The audience is a hypothesis

"People asking varied questions" describes a motivation, task or context for which this direction tends to be chosen. It was not measured, and it says nothing about any age, gender or other group.

UX risk and how to reduce it

Risk: Confusing mix of channels and data.

Mitigation and accessibility: Typed contracts and a text alternative.

How to check: Test with realistic scenarios: task success, accuracy, reversal, auditability, understanding of what the system does on its own, and a human's ability to step in.

Evidence

Emerging: exploratory work, with no causal validation for this style.

The sources below support the principles that apply to this term. None of them shows that this style works better or is preferred.

Lee 2025 (Preliminary): Definition of generative UI.
Luera 2025 (Supported): Interaction patterns for generative AI.
Chen 2025 (Preliminary): LLM-generated user interfaces.

Heuristic scales

Relative editorial judgements from 1 to 5. They are not results of a study and not a measure of quality.

Visual complexity heuristic: 4 / 5 (1 is few visual elements, 5 is many layered elements)
Expressiveness heuristic: 4 / 5 (1 is restrained, neutral, 5 is highly expressive)
Information density heuristic: 4 / 5 (1 is sparse, 5 is dense)

In our library

No design-style recipe in @codeatlasdev/ui implements this term yet. You can still build it from the --ca-* tokens: see Theming.

Prompt for your AI

Paste this into a coding agent. It carries the trade-offs, and asks for a baseline to compare against.

Design this interface in the "Multimodal Generative UI" direction (future / ai, dynamic composition): Combines on-demand interface with voice, charts, forms and documents.
Visual and interaction traits to use: chat, charts, actions.
Main UX risk: Confusing mix of channels and data. Mitigate it: Typed contracts and a text alternative.
Keep it accessible to WCAG 2.2 AA: contrast, visible focus, keyboard operation, 24 px minimum targets, reflow at 320 CSS px and reduced motion.
Keep the content, hierarchy and behavior the same as a plain baseline so the two can be compared. Do not claim the style performs better; the choice should be validated with users.

Limits

The atlas is a broad, living catalogue, not every style that exists. No authority enumerates them, and the classes (movement, aesthetic, design system, layout, pattern, principle) overlap.
The 1 to 5 scores are relative editorial judgements. They are not results of a study and not a measure of quality.
An audience is a hypothesis about motivation, task and context. It was never measured and says nothing about any age, gender or other group.
The sources support underlying principles (visual complexity, legibility, grouping, accessibility, platform conventions). None shows that a particular style works better or is preferred.
The 2027 to 2030 trend is an editorial projection, not a forecast.
Terms in the Future / AI category rest on exploratory work, some of it preprints, with no causal validation.

Last reviewed 2026-10-25.

References

Lee (2025). Towards a Working Definition of Designing Generative User Interfaces. arXiv preprint 2505.15049.Preliminary
Supports: A multi-method qualitative study (systematic review of 127 publications, 18 expert interviews, 12 case studies) that proposes a working definition of generative UI as an iterative, co-creative process.
Does not support: Effectiveness of generative UI. It is a definitional, qualitative study and a preprint.
Checked 2026-10-25.
Luera, Rossi, Siu, Dernoncourt, Yu, Kim, Zhang, Chen, Salehy, Lipka, Basu, Mathur, Zhao (2025). Survey on User Interface Design and Interactions for Generative AI Applications. Foundations and Trends in Human-Computer Interaction.Supported
Supports: A survey that presents taxonomies of how a human interacts with AI and of the user-interaction patterns designed for a variety of generative AI use cases, as a reference for designers and developers. It focuses primarily on user-guided interactions, those the user initiates.
Does not support: That any pattern works better, or that a generative interface should act without the user's direction: the survey leaves out implicit signals from the user and reports no outcome comparison. The abstract read is the arXiv version of the same survey (earlier title); the published monograph's own text was not reachable.
Checked 2026-10-25.
Chen, Zhang, Zhang and others (2025). Generative Interfaces for Language Models. arXiv preprint 2508.19227.Preliminary
Supports: Proposes a paradigm in which language models respond to queries by generating task-specific user interfaces, with a framework of structured interface representations and iterative refinement, and reports an evaluation.
Does not support: That generated interfaces beat conversation in general. The abstract read here is a proposal and a preprint; its evaluation was not examined.
Checked 2026-10-25.