I am reporting what I believe is a serious information-architecture and usability problem in ChatGPT Projects.
Goal: I need ChatGPT Projects to support stable, user-controlled identities for long-running professional conversations.
Current workaround: I am forced to maintain external migration records, checkpoints and searchable identifiers inside chats because conversation titles cannot be relied upon as stable identifiers.
Ideal outcome: Once a user manually renames a chat, that title should remain unchanged unless the user explicitly changes it again. Ideally, ChatGPT Projects should also support permanent conversation aliases, active/current conversation markers and explicit predecessor/successor relationships.
Please do not treat this as a request for help finding one lost conversation. This is a systemic product and information-architecture issue affecting long-running professional workflows.
I use ChatGPT extensively for complex scientific, research and business work. A single Project may contain many conversations. Some are independent workstreams, while others are direct continuations or successors of previous chats. We maintain continuity through migration packages, checkpoints, explicit predecessor/successor relationships and carefully named conversations.
For example, I may deliberately maintain a central sequence such as:
007_2026_08_10
008_2026_08_13
009_2026_08_15
These are not decorative titles. They identify successive generations of a long-running workstream.
The problem is that ChatGPT may dynamically change conversation titles, and there appears to be no way for the user to permanently lock a title.
For casual use, automatic retitling may sometimes be convenient. For serious professional use, it can destroy the human-readable structure of a Project.
If ChatGPT changes deliberately assigned titles into descriptive AI-generated titles, I can no longer reliably determine from the Project interface:
- which conversation is the current successor;
- which conversation is obsolete;
- which belongs to a parallel workstream;
- what the chronological sequence is;
- where important decisions were made;
- which conversation should be continued.
The conversations may technically still exist, so no data has necessarily been deleted. However, the user’s information architecture has been damaged.
This becomes particularly problematic because ChatGPT Projects currently have a largely flat structure: Project → conversations. There is no full hierarchical structure with folders, subprojects, workstreams or explicit successor chains that could compensate for unstable titles.
The resulting combination is:
flat Project structure + dynamically mutable chat titles + no permanent user-controlled conversation identity
For long-running professional work, that combination is not reliable.
Search does not solve the underlying problem.
Search is useful when the user already knows a distinctive phrase or identifier contained inside the correct conversation. But a professional information system should not require users to remember arbitrary phrases or invent hidden search tokens merely because the application may later change the names of their work objects.
The sidebar title should not be treated as disposable AI-generated metadata once the user has deliberately assigned a name to it.
Requested changes
1. Never automatically overwrite a manually assigned title
This is the most important and simplest fix.
Once a user manually renames a conversation, that title should become user-controlled metadata.
ChatGPT may suggest another title, but it should never silently replace a title deliberately assigned by the user.
The minimum viable rule should be:
manual title → permanent unless the user explicitly changes it
This alone would solve a large part of the problem.
2. Add a “Lock title” option
For example:
Rename
Lock title
Once locked, the title remains unchanged until the user explicitly unlocks it.
This would preserve automatic title generation for casual users while giving professional users deterministic control.
3. Provide a permanent user-defined conversation alias or stable ID
Every conversation clearly has an internal technical identity.
Users should be able to assign a permanent human-readable alias such as:
200-MP-009
The display title could still exist independently, but the permanent alias should never change.
4. Add an “Active / Current conversation” function inside Projects
A Project should allow users to mark a conversation as:
Active main conversation
or:
Current conversation for this workstream
Then a user could simply select:
Continue active conversation
instead of reconstructing which of many related chats is current.
5. Support explicit predecessor/successor relationships
Long conversations eventually need to continue in new chats.
ChatGPT should provide something like:
Continue this conversation in a new chat
and visibly preserve the relationship:
Chat 007 → Chat 008 → Chat 009
The successor should know its predecessor, and the predecessor should visibly point to its successor.
This is much more reliable than expecting users to infer continuity from titles.
6. Improve Project hierarchy and navigation
For serious work, Projects would benefit greatly from at least some of the following:
- folders or subprojects;
- workstreams;
- branches;
- chronological sorting independent of title;
- clear creation timestamps;
- clear last-activity timestamps;
- pinned conversations;
- active/current conversations;
- archived or retired conversations;
- explicit main-thread versus branch distinction.
At present, a complex Project can eventually become a flat collection of conversations whose titles are not guaranteed to remain stable.
Why this matters
ChatGPT is increasingly capable of supporting sophisticated professional work: scientific research, business planning, writing, coding, long-term analysis and other persistent workflows.
Those use cases require basic information-management properties:
stable identity, chronology, provenance and user-controlled organization.
A dynamically generated title is useful when it is merely a suggestion.
It becomes harmful when the system treats the title as disposable while the user treats it as the stable name of a persistent work object.
The governing principle should be:
AI-generated metadata may assist the user, but it should never silently overwrite user-defined organizational metadata.
This change would not harm casual users.
Automatic titles could remain the default. Users who never rename chats would experience exactly the same product.
But once a user deliberately names a conversation, the system should respect that decision.
I am heavily invested in ChatGPT as a long-term professional working environment. That is precisely why this issue is concerning: the underlying model capabilities are increasingly suitable for sophisticated professional workflows, while a very basic information-management behaviour can undermine the usability of the entire system.
I have also submitted this issue privately through OpenAI Support, where it has been escalated to a support specialist. I am posting it here as a public product/UX feature request because I believe the issue affects many users who rely on ChatGPT Projects for serious long-term work.
Please consider routing this feedback to the relevant ChatGPT Projects / product / UX / engineering team.
At an absolute minimum, I strongly request:
Once a user manually renames a conversation, that title should remain unchanged unless the user explicitly changes it again.