I would like to propose a feature enhancement for ChatGPT conversations that allows users to rollback or truncate a conversation from a selected message onward.
Problem Statement
During ongoing conversations, especially on complex or exploratory topics, users often gain clarity progressively. As understanding evolves, it becomes apparent that:
-
The conversation may have deviated from the intended direction
-
Earlier assumptions or prompts were suboptimal
-
A different line of inquiry would yield better outcomes
Currently, the only way to correct this is to manually redirect the conversation through new prompts. However:
-
The model still retains the full prior context
-
Earlier misdirection continues to influence responses
-
There is no clean way to “branch” or restart from a better point
Example Scenario
-
A user is in middle of current conversation of scientific topic (lets say at prompt no. 21).
-
User realizes that the discussion has gone off-track from prompt no. 15
-
The user wants to:
-
Remove messages 16–21
-
Restart the conversation from message 15 with a revised direction
-
At present, this is not possible without starting an entirely new chat or working around residual context.
Proposed Solution
Introduce a “Delete Forward / Rollback” option for each message in the conversation.
Feature Behaviour:
-
Add a button alongside existing options (e.g., Copy, Edit) for each message that AI or the user provides.
-
Label suggestion:
-
“Rollback to Here”
-
or “Delete all messages from here”
-
On click:
-
small message box appears asking to confirm to delete all further messages from this prompt? with a Yes/No button. (alerts user that this action cannot be reverted).
-
If clicked YES: All subsequent messages (both user and AI) are deleted
-
The conversation state resets to that selected point
-
The user can continue the conversation from there with fresh context
Benefits
-
Enables clean conversational branching
-
Improves learning efficiency and exploration quality
-
Reduces context pollution from irrelevant or incorrect prompts
-
Enhances user control and precision in prompt engineering
-
Particularly valuable for:
-
Research workflows
-
Technical problem-solving
-
Iterative ideation
-
Optional Enhancements
-
Allow users to create multiple branches from a single message (versioning system)
-
Provide a visual indicator of rollback points
Conclusion
This feature would significantly improve the usability of ChatGPT for advanced and iterative use cases. It aligns well with how users naturally think—refining direction as clarity improves—and would help unlock more effective interactions with AI.