Feature Request: Let AI Organize My Conversations by Topic, Not Just Chronology
I have become a frequent ChatGPT user, and as my conversation history has grown, I have encountered a usability problem that I believe will become increasingly important as people use ChatGPT for ongoing work and decision-making.
The current model treats each chat primarily as a separate conversation in a chronological history. Search helps, but it still requires me to remember which conversation contained the information I am looking for—or at least remember the right words to search for.
In practice, that is not how I use ChatGPT.
I may discuss the same underlying subject in numerous separate conversations over weeks or months. For example, I might have one conversation about whether to sell a particular real-estate property, another about its rental economics, another about the surrounding real-estate market, another about tax implications, and another about communications with the listing agent.
To me, those are not five unrelated conversations. They are all part of one ongoing topic or project.
ChatGPT itself is capable of understanding that relationship. So why should the user have to organize it manually?
Proposed Feature: AI-Organized Topics
I would like ChatGPT to automatically recognize related conversations and organize them into persistent topics or projects based on their semantic content.
For example, instead of only seeing a chronological list of hundreds of chats, I might see AI-generated topics such as:
Real Estate – Property A
LifeSprings / Investment Opportunities
Podcast
Travel Planning
Family
Technology
A conversation could belong to more than one topic when appropriate, and users could rename, combine, separate or correct the AI-generated classifications.
But the important point is this: I should not have to manually create and maintain the filing system. The AI should do that for me.
Take It One Step Further: Topic-Level Intelligence
Ideally, opening a topic would provide more than a folder containing old conversations.
ChatGPT could maintain an AI-generated summary of the topic showing:
-
Important facts established across conversations
-
Decisions that have already been made
-
Important changes over time
-
Outstanding questions or decisions
-
Links to the conversations where the underlying discussions occurred
I could then ask:
“Show me everything we’ve discussed about whether to sell or rent Property A.”
Or:
“Where did we discuss the actual rental income we received?”
ChatGPT could answer across my conversation history and point me to the original conversations when needed.
Why Search Alone Doesn’t Solve the Problem
Search assumes that I remember enough about the old conversation to know what to search for.
As conversation history becomes larger, the real problem isn’t simply finding a particular word or phrase. It is remembering where an idea, analysis, decision or piece of information fits within months or years of interaction with ChatGPT.
AI is uniquely suited to solve that problem.
Traditional software would require users to create folders, tags and naming conventions. But an AI system already understands what conversations are about and how concepts relate to one another.
The goal should therefore be:
Don’t make users organize their AI conversations. Let the AI understand how their conversations relate to one another and organize them for them.
I believe this becomes increasingly important as ChatGPT evolves from something people occasionally ask questions of into something they use continuously for business, personal projects, research, planning and decision-making.
One related suggestion: after a conversation like this one, it would also be useful if I could simply tell ChatGPT, “Submit this as product feedback to OpenAI,” rather than leaving ChatGPT and manually submitting the idea somewhere else.