openai / openai/codex

[ChatGPT desktop] “Add to chat” follow-up auto-scrolls away from the selected passage

Open
#41,391 0 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

app enhancement
Dominant language
Rust
Stars
125k
Forks
19.4k
PR merge metrics
PR metrics pending

Description

What variant of the app are you using?

The ChatGPT chat surface in the unified ChatGPT/Codex desktop app on macOS. This report is about regular ChatGPT chats, not Codex tasks.

App version

ChatGPT 26.820.60940 (build 7119)

Platform

macOS 26.5.2 (25F84), Apple Silicon (arm64)

What issue are you seeing?

While reading an earlier part of a long ChatGPT answer, I can select a passage, choose Add to chat, and write a follow-up question about that selection.

When I send the follow-up, the app automatically scrolls to the bottom of the conversation. This removes the selected passage from view and makes me lose the exact place I was reading.

The selected-text annotation also does not provide an obvious way to return to its source passage afterward. Because the follow-up was created from that specific passage, the annotation should act as a navigation anchor rather than only as context sent to the model.

Steps to reproduce
  1. Open a ChatGPT conversation containing a long assistant response.
  2. Scroll to a passage that is not currently at the bottom of the conversation.
  3. Select part of the assistant response.
  4. Choose Add to chat.
  5. Type a question about the selected passage and send it.
  6. Observe that the conversation automatically scrolls to the newest message at the bottom.
  7. Try to use the submitted annotation to return to the original selected passage.
Actual behavior
  • Sending the annotated follow-up jumps the viewport to the bottom of the chat.
  • The user loses the passage and reading position they were reviewing.
  • The annotation is not an obvious clickable backlink to the selected source text.
Expected behavior

When a follow-up is created from selected text, the app should preserve the user's reading position instead of forcing the viewport to the bottom.

A non-disruptive Jump to latest or New response control could let the user move to the new turn when ready.

In addition, clicking the submitted annotation should scroll back to the original passage and briefly highlight it.

Ideally, the app should support both behaviors:

  1. preserve the reading position when the annotated follow-up is sent; and
  2. make the annotation a bidirectional link to the source passage.
Why this matters

Selecting text and asking a question about it is an in-context reading workflow. Auto-scrolling away at submission breaks that context, especially in long research, technical, or analytical answers, and forces the user to manually search for the passage again.

Related issues

These are related but do not appear to cover this selection-specific behavior:

  • #39928 — show annotation contents inline in chat history
  • #34057 — add in-conversation previews and quick navigation to ChatGPT chats
  • #33770 — add prompt navigation for long chats to the macOS desktop app

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

The issue names no repository files, tests, or entry points; first confirm whether the ChatGPT desktop behavior is implemented in this repository. Done would preserve the reading position after sending an annotated follow-up and make the submitted annotation return to and highlight its source passage.

Written by the indexing model from the issue text.

Assessment

Tech stack
macos
Domain
desktop
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Active
Clarity
Mostly clear
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.