Feature Request: Proactive Memory — Contextual Follow-ups Based on User Memory
The idea
I would like to suggest a new capability that combines Memory + Tasks + conversational context to allow ChatGPT to proactively follow up on things users have previously mentioned.
ChatGPT already has Memory, which allows it to remember information about a user across conversations, and Tasks can remind users about things they explicitly schedule.
The missing piece, in my opinion, is contextual follow-up.
Instead of only remembering something or waiting for the user to create a reminder, ChatGPT could recognize that something the user mentioned may be worth revisiting later and, with the user’s explicit permission, initiate a follow-up conversation when the timing makes sense.
Example
A user says:
“I’m studying for an exam in November.”
ChatGPT could respond:
“Would you like me to keep track of this and check in with you as the exam approaches?”
If the user agrees, ChatGPT could later initiate:
“You mentioned that you’re preparing for an exam in November. How is your preparation going? Would you like me to help you review the subjects you still need to study?”
This would be different from simply setting a reminder for a specific date.
The follow-up would be generated from the previous conversation and the current context.
Other examples
Travel
User:
“I’m traveling in December.”
Later:
“Your trip is getting closer. Have you already arranged your accommodation and transportation?”
If the user previously mentioned that accommodation was still unresolved, the follow-up could specifically ask about that.
Long-term projects
User:
“I’m writing a fantasy novel and I’m currently developing the magic system.”
Several weeks later:
“How is your novel progressing? Last time we discussed it, you were developing the magic system. Would you like to continue from there?”
Studying
User:
“I’m preparing for a public-service exam and I’m studying 30 minutes every night.”
ChatGPT could periodically check in:
“How is your study routine going? Have you been able to maintain the 30 minutes per night?”
It could then offer assistance based on what was previously discussed.
Health-related situations
If a user says they have started a treatment and wants to observe how things develop, ChatGPT could ask later:
“You mentioned that you wanted to see how things developed over the next few days. How have things been?”
For health-related subjects, this should remain a supportive follow-up rather than medical decision-making, with appropriate safety safeguards.
Why this is different from Tasks
I think this distinction is important.
Traditional Task:
“Remind me on October 15 to study.”
Proactive Memory:
“I remember that you’re preparing for an exam. I’ll check in with you as the exam gets closer and ask how your preparation is going.”
The first is a scheduled reminder.
The second is continuity and contextual assistance.
The user doesn’t necessarily know in advance exactly when a follow-up will be useful. ChatGPT could determine an appropriate moment based on the information and timeline the user provided.
User control and privacy
Because proactive communication could become intrusive, this should be strictly opt-in.
For example, when ChatGPT detects a potential opportunity:
“This sounds like something you might want to follow up on later. Would you like me to keep track of it?”
Users could also have settings such as:
- Enable/disable Proactive Memory
- Choose which categories can generate follow-ups
- View active/upcoming follow-ups
- Pause or delete individual follow-ups
- Control how frequently ChatGPT can initiate conversations
- Prevent follow-ups for sensitive topics
The user should also be able to see why a follow-up was created and what previous information it is based on.
A possible “Follow-ups” section
ChatGPT could have a simple area showing things it has been authorized to follow up on:
Upcoming follow-ups
Exam preparation — November
Trip preparation — December
Novel project — Ongoing
Each could have options such as:
Edit · Pause · Delete · Don’t follow up again
The larger opportunity
I think this could be a natural evolution of ChatGPT’s Memory.
Currently:
Memory:
“I remember what you told me.”
Tasks:
“I’ll remind you when you tell me to.”
Proactive Memory:
“I remember what you told me, I understand that it may become relevant again, and—if you’ve authorized me—I can check in when it makes sense.”
This could make ChatGPT feel much more like a continuous personal assistant rather than a system that only responds when the user remembers to come back.
The goal wouldn’t be for ChatGPT to constantly send notifications.
The goal would be for it to recognize:
“This is something the user may want help remembering, monitoring, or revisiting later.”
And, with the user’s permission, actually follow through.
I believe this could be a particularly powerful use of Memory because it would transform stored context from something mostly passive into something that enables continuity, accountability, and meaningful long-term assistance.
Thank you for considering the idea.