Feature Request: Let Users Approve, Remember & Lock Specific AI-Generated Ideas
The problem
ChatGPT conversations often become collaborative projects.
For example:
-
A designer is developing a campaign.
-
A writer is refining a story.
-
A developer is building a product.
-
A researcher is developing a hypothesis.
-
A user is planning something over many messages.
During these conversations, ChatGPT may generate 10–20 useful ideas in a single response.
The user may strongly agree with only 3 or 4 of them.
Today, there is no simple way to tell ChatGPT:
“This specific point is approved. Keep using it.”
The user has to scroll back, copy the relevant point, and repeat it in a later message.
As conversations become longer, this creates unnecessary friction and increases the chance that previously approved decisions get lost.
My proposed solution
Add lightweight interaction controls to important parts of ChatGPT’s responses.
For example:
Like
“I like this idea.”
✓ Agree
“This is an approved direction for the current conversation/project.”
Remember
“Remember this as a user preference.”
Lock
“This is now a fixed project constraint. Don’t change it unless I explicitly unlock it.”
Reject
“Don’t continue developing this direction.”
The user could apply these controls to:
-
A paragraph
-
A bullet point
-
A specific idea
-
A recommendation
-
A creative direction
-
A decision
-
Potentially even a sentence or selected portion of a response
The important distinction
I think Agree, Remember and Lock should behave differently.
✓ Agree
Temporary/project-level approval.
Example:
“Use photorealistic imagery.”
ChatGPT understands:
This direction has been approved for this project.
Remember
Long-term user preference.
Example:
“I generally prefer minimal, premium visual design.”
If the user chooses “Remember”, this could become part of their saved preferences, subject to the user’s existing Memory settings and controls.
Lock
A hard constraint for the current project.
Example:
“The logo must not be changed.”
ChatGPT should treat this as a fixed constraint until the user explicitly unlocks it.
An example from a real creative workflow
I recently used ChatGPT to develop an internal advertising-agency design contest.
ChatGPT generated several creative recommendations.
I approved specific points such as:
✓ Use the employee’s actual photograph for the introduction.
✓ Keep the visual treatment photorealistic and premium.
✓ Exaggerate the person’s real talents rather than inventing unrelated characteristics.
✓ Keep the tone approximately 50% uplifting and 50% humorous.
✓ Create a character reference before generating the final AI scenes.
✓ Allow minor variations between employees.
Today, I have to manually tell ChatGPT which of these decisions I approved.
With the proposed feature, I could simply click:
✓ Agree
or
Lock
on the relevant recommendations.
ChatGPT could then maintain a lightweight project decision layer in the background.
Even better: a Project Decision Panel
For long-running projects, ChatGPT could optionally show something like:
PROJECT DECISIONS
Approved
-
Photorealistic visual style
-
50% humour / 50% uplifting
-
Real employee photograph for introduction
Locked
-
Carousel format: 1080 × 1350
-
Six-slide structure
-
Slide 1 introduces the real employee
Rejected
- Alternate-universe concept
User preferences
-
Premium visual design
-
Minimal unnecessary text
This would become a living brief that evolves naturally during the conversation.
Why I think this would be valuable
This could significantly improve:
-
Creative direction
-
Design workflows
-
Writing projects
-
Coding projects
-
Research
-
Business planning
-
Long conversations
-
Projects with multiple iterations
-
Collaboration between humans and AI
It would also reduce repetitive prompting.
Instead of:
“Remember what we decided earlier…”
the user could simply approve the decision when it is made.
A possible future interaction
Imagine ChatGPT responds:
Here are 6 possible creative directions…
The user clicks:
✓ Agree on #2
✓ Agree on #5
Lock on #5
Reject on #1
ChatGPT automatically understands:
Develop #2 and #5 further.
#5 is locked and should remain part of the final direction.
Don’t return to #1 unless explicitly requested.
That feels much closer to how a human creative collaboration actually works.
One more idea: “Approve all”
When ChatGPT gives a structured response with several recommendations:
Approve all ✓
Reject all ✕
This would make the feature especially useful for creative and professional workflows.
Privacy and user control
This should be completely user-controlled.
The UI should clearly distinguish between:
Current conversation/project decisions
and
Long-term memory/preferences.
Nothing should become long-term memory simply because the user clicked “Like” or “Agree”.
The user should explicitly choose Remember for persistent preferences.
The bigger opportunity
I don’t see this simply as adding more buttons.
I see it as adding a new layer to ChatGPT:
Conversation → Decisions → Memory → Action
Today, ChatGPT is very good at generating ideas.
The next step could be making it much better at understanding which ideas the human has actually approved.
That could make long-term human–AI collaboration feel dramatically more natural.
Thank you for considering this.
Rajnikanth