Until recently, ChatGPT let you edit any message in a conversation and regenerate responses from any point in the thread. That's no longer the case.
OpenAI has updated ChatGPT so that you can only edit your most recent message and retry only the most recent response. The change is documented on OpenAI's help page for GPT-5.3 and GPT-5.4, but there was no blog post or in-app notification announcing it.
For casual users, this probably changes nothing. But for anyone who treats conversation branching as a workflow tool, testing different prompts mid-conversation, exploring alternative response paths, or backing up to fix earlier instructions, this is a real downgrade.
What Power Users Lose
The old behavior let you treat a ChatGPT conversation like a tree: branch off at any point, try different approaches, compare results. Common use cases that no longer work:
- Prompt iteration: editing an earlier message to refine instructions without losing later context
- A/B testing responses: regenerating from a specific point to compare different outputs
- Error recovery: jumping back several messages when a conversation goes off track
Now your only options are editing the last thing you typed or retrying the last thing ChatGPT said. To change something from five messages ago, you need to start a new conversation or manually copy-paste your revised prompt.
The timing suggests this is a technical constraint of the newer GPT-5.3 and GPT-5.4 models rather than a deliberate product decision, though OpenAI hasn't explained the reasoning. The silent rollout is the frustrating part. When you change how a core feature works, telling your users should not be optional.