Introduction
Once an AI generates a response, what can users do with it? Across 12 systems, we found consistent patterns for core actions (copy, feedback) but significant divergence in how systems handle response refinement, follow-up suggestions, and content acceptance workflows.
Key finding: Standalone chat systems use compact 4-6 icon action bars, while integrated systems (document editors, app builders) require richer control sets with Accept/Discard workflows. Stop generation is universal (12 of 12), but version control remains rare (2 of 12).
Action Bar Patterns
| Pattern | Count | % | Systems |
|---|---|---|---|
| 4-5 icon compact bar | 6 | 50% | Claude, Gemini, GitHub Copilot, v0, Notion AI (sidebar), Retool AI |
| 6+ icon expanded bar | 4 | 33% | ChatGPT, Perplexity, Microsoft Copilot, Atlassian Rovo |
| Accept/Discard workflow | 4 | 33% | Notion AI, Google Docs AI, Atlassian Rovo, Airtable |
| Refine/Edit dropdown | 3 | 25% |