Hi OpenAI team,
I’d like to suggest paragraph-level emoji reactions in ChatGPT responses.
Right now, feedback/reactions apply to the entire assistant message. For longer answers, it would be much more useful if each paragraph had a small reaction control such as 
.
Why this would help:
- A user can mark exactly which paragraph was useful, unclear, funny, wrong, or especially important.
- The reactions could optionally become structured feedback/memory metadata: paragraph ID → reaction → date/topic.
- This could make long conversations easier to navigate later, for example finding paragraphs marked or unresolved .
- Global actions such as copy, share, read aloud, etc. could remain at the bottom of the full message.
An optional “remember this” signal per paragraph could also help users explicitly mark information they want carried forward.
Thank you for considering it.