Long answers stream linearly — users wait without knowing if the answer will be three bullets or three pages.
← Back to patterns
Google
Visual & Generative
EmergingOutput Skeleton
Show the shape before filling it
Instead of streaming a long answer character-by-character, render the structure first — headings, section count, expected formats — then fill the slots in. Users know within a second whether they will get a 3-bullet TL;DR or a 1,200-word essay, and can start orienting before the content arrives.
Metaphor
Wireframe
Interaction
Schema-First Reveal
Use case
Reports, Briefs, Long Structured Outputs, Code Files
Failure mode
Shape Hidden Until End
Control mechanisms
Pre-rendered Outline · Section Placeholders · Staggered Fill
Reference examples
ChatGPT Deep Research · Gemini Deep Research · Claude Artifacts (skeleton behavior)
Concept Frame
Predictability is a layout, not a promise.
Streaming text feels fast because something is moving — but it hides the answer's shape until the very last token. Was it three bullets or three pages? You only find out when the cursor stops. Output Skeleton flips the order: within a few hundred milliseconds the user sees the headings, the table placeholder, and the length envelope. From that moment, the wait is no longer anxious — it is scoped.
Interactive Example
The live side-by-side is the comparison.
This prototype already contains the contrast between chat and the pattern response. It stands as the primary comparison surface for this page.
You waited for the shape. It should arrive first.
Long answers stream linearly, so the reader sits through a character-by-character reveal with no idea whether the payoff is three bullets or three pages. Chat treats output as a flat stream, which means the shape you were promised only resolves at the very end — exactly when it's least useful for orienting. The structure is the cheapest part of the answer, and it's delivered last.
Output Skeleton inverts the order: render headings, section count, and expected formats first, then fill the slots — often in parallel. Within a second the user knows the envelope and can skip, expand, collapse, or reject the structure before content commits to it. Orientation stops being a reward for waiting and becomes the opening move.
"Show the wireframe before the copy; a reader who knows the shape can start thinking before the words arrive."
DOPE Lens
How this pattern improves AI interaction quality.
Observability
The shape of the answer is visible immediately, before any content arrives
Directability
Skip ahead, expand or collapse sections, reject the structure before content fills
Predictability
Section count, format, and length envelope are declared up front
Explainability
Each placeholder declares what kind of content will land there (paragraph, list, table)
Fit
Where it earns its place.
Use it
when
when
Long outputs have predictable structure and users benefit from seeing the shape before content arrives.
Avoid
when
when
Outputs are short or freeform enough that a structural preview would be heavier than the content itself.
In the wild
Where this pattern already ships.
The pattern is shipping but still maturing across products. Entries below show how different teams have approached it; expect variation as the space settles.
ChatGPT Deep Research
OpenAI
Proposes a research plan card before execution; the user reviews and modifies the outline, then the agent runs. The plan-first contract made explicit.
Visit OpenAI ↗
Gemini Deep Research
Apr 2026 launch on Gemini 3.1 Pro made the editable research plan the hero UX — the structure of the answer is shown and shaped before any content arrives.
Visit Google ↗
Claude Artifacts (skeleton behavior)
Anthropic
Caveat — implicit, not a named feature. For long structured documents Claude often drafts headings and sections first, then fills them. Pattern visible in behavior, not branding.
Visit Anthropic ↗
Research Frame
Why this pattern exists, and where it breaks.
Render the structure first (sections, formats, lengths) and fill content into the placeholders, often in parallel.
Chat treats output as a flat character stream; the user must wait until the end to learn the shape they were promised.
Skeletons that overpromise (more sections than the model can fill well) feel like a bait-and-switch.
Common questions
What people ask about this pattern.
When should I use the Output Skeleton pattern?
+
Long outputs have predictable structure and users benefit from seeing the shape before content arrives.
When should I avoid the Output Skeleton pattern?
+
Outputs are short or freeform enough that a structural preview would be heavier than the content itself.
What problem does Output Skeleton solve?
+
Long answers stream linearly — users wait without knowing if the answer will be three bullets or three pages.
Why is chat the wrong fit for this?
+
Chat treats output as a flat character stream; the user must wait until the end to learn the shape they were promised.
Related Patterns
Curated connections, not just neighbors.
Hand-picked relationships — what this pattern pairs with in practice, what it replaces, what it builds on, and what builds on it.
Often paired with
Progressive Disclosure
Output Skeleton lets you open only the answer sections you need; Progressive Disclosure does the same for advanced controls.
Often paired with
Generative UI
A skeleton is the contract before Generative UI fills it in — structure first, components later.
Often paired with
Intent Preview
Intent Preview shows the steps an agent will take; Output Skeleton shows the shape of what it will hand back.