ChatGPTOutput & Refinement
ChatGPT renders structured answers in the thread with reasoning activity, version navigation, feedback, sharing, and selection-based refinement into writing blocks.
Structured answer with collapsed Thought for 5s reasoning label
What's happening
ChatGPT keeps most output in the message stream. Reasoning models expose a collapsible Thought label that opens an Activity panel. Each reply gets version navigation, feedback, share, and overflow actions. Users can refine a text selection or promote it to a writing block.
Patterns
Regenerate with version history; Ask ChatGPT or Start writing on selections
Thumbs up confirms in one click; thumbs down opens categorized feedback
UX Insights
- •Reasoning stays out of the main answer until the user expands Activity
- •Version pagination (4/4) preserves prior drafts without branching the whole chat
- •Selection refinement targets a passage instead of regenerating the full reply
- •Writing blocks bridge in-thread answers and editable deliverables
Design Decisions
ChatGPT spreads refinement across regenerate, selection actions, and writing blocks rather than one contextual menu. Feedback asymmetry , quick thumbs up, structured thumbs down , matches triage needs while keeping positive signal low-friction.
More case studies
ClaudeOutput & Refinement
Claude keeps short answers in the message stream with per-message actions, non-destructive Try again, and version navigation.
Google GeminiOutput & Refinement
Gemini renders long-form output in-thread with inline source chips, comparison tables, section export, follow-up chips, refinement menus, Workspace handoffs, Maps split-pane, and a sources sidebar.
PerplexityOutput & Refinement
Perplexity renders search answers with Answer/Links/Images tabs, research steps, follow-ups, export, rewrite, and selection refinement.
ClaudeArtifacts
Substantial work opens as an artifact card, then expands into a split pane with live preview, code view, and optional publish.