Are ChatGPT Custom Instructions the same as Memory?
No. Custom Instructions are explicit guidance that you write and manage. ChatGPT Memory is a separate personalization feature that can use relevant information from conversations.
Comparison
ChatGPT Custom Instructions vs thredly
Custom Instructions are for standing rules. thredly is for the specific context of a conversation that has already accumulated work.
ChatGPT feature details verified against official OpenAI Help Center documentation on 2 September 2026.
The difference in one sentence
ChatGPT Custom Instructions are standing guidance that you ask ChatGPT to consider in its responses, such as your preferred tone, format, background, or recurring rules. OpenAI currently says they are available on all ChatGPT plans across web, desktop, iOS, and Android, and that updates apply across chats when customization is enabled.
A thread-specific continuation handover is different: it is a reviewed snapshot of one conversation's working state, including its objective, active constraints, decisions, completed work, latest output, unresolved issues, and exact next action.
ChatGPT Custom Instructions versus a conversation handover
| Decision factor | ChatGPT Custom Instructions | Thread-specific handover |
|---|---|---|
| Purpose | Shape ChatGPT's recurring behaviour and responses. | Continue one specific piece of work in a fresh conversation. |
| Scope | Broad guidance used across chats where customization is enabled. | The working context from one chosen conversation. |
| Persistence | Remains in your personalization settings until you edit, disable, or delete it. | Used when you paste it into a chosen chat; it does not persist or sync automatically. |
| Task-specific decisions | Only present if you manually put them into a standing setting, where they can become stale. | Records the decisions that still govern this task and, where useful, why they were made. |
| Current progress | Does not reconstruct the current state of an old thread by itself. | States what is finished, in progress, blocked, or waiting for review. |
| Latest draft or output | Does not automatically carry the latest approved work from a particular conversation. | Can include the latest approved result or a concise description of where to find it. |
| Unresolved actions | Not designed as a live task checklist for one thread. | Calls out open questions, risks, and the next action. |
| User review | You manage the standing instructions in ChatGPT settings. | You review the generated or manual handover before using it and remove errors or stale context. |
| Best use case | Stable preferences, style, role, and reusable response rules. | A long or unreliable conversation that needs to continue with its task state intact. |
When Custom Instructions are enough
- You want every response to use UK English, a concise tone, or a consistent format.
- You want ChatGPT to understand a recurring role or stable background that applies beyond one task.
- You have reusable rules such as asking for assumptions, using accessible language, or presenting recommendations in a table.
- You do not need the decisions, latest draft, or current progress from one specific conversation.
When a thread-specific handover is better
- A long AI chat has become slow, inconsistent, or difficult to continue reliably.
- The next chat needs the objective, active constraints, decisions, work completed, and exact next action.
- The latest draft or output matters, while older rejected approaches should stay behind.
- You want a visible summary that you can review, edit, and paste into the new conversation.
When using both is sensible
Custom Instructions and a handover can be complementary. Keep durable preferences in Custom Instructions, then place the reviewed, task-specific handover at the start of the fresh chat. One layer tells ChatGPT how you generally want it to respond; the other explains what this particular job is and where the work has reached.
For example, your standing instructions might say: 'Use UK English, be concise, and label assumptions.' The handover might say: 'Objective: finish the launch brief. Decisions: lead with the time-saving benefit and avoid the rejected automation claim. Current progress: sections one to four are approved. Latest output: revised positioning paragraph. Next action: draft the risks section.'
If you work inside a ChatGPT Project, project instructions are another standing layer. OpenAI currently says they apply only inside that project and override global Custom Instructions there; they still serve a different purpose from a reviewed snapshot of one thread's latest working state.
Custom Instructions, Memory, and conversation summaries are different
Custom Instructions are explicit guidance that you write and manage. ChatGPT Memory is a separate personalization feature that may use information from conversations. A conversation summary or handover is a deliberately selected account of one thread's task state. These features can overlap in what the model sees, but they are not interchangeable.
Changing Custom Instructions does not rewrite the messages already visible in your chat history. And Custom Instructions do not, by themselves, recreate every decision, draft, or unresolved action from a previous conversation. For a closer feature comparison, see ChatGPT Memory versus a thread handover and the ChatGPT Memory guide.
Before and after: continuing a real piece of work
Before: a strategy thread contains an approved audience, two rejected positioning routes, a working outline, the latest introduction, and one unresolved pricing question. Your Custom Instructions say to write in a direct, professional style and show recommendations in tables.
Fresh chat with Custom Instructions only: ChatGPT can follow your preferred style, but it has not been given the approved audience, rejected routes, current outline, latest introduction, or pricing question.
Fresh chat with both: the standing instructions shape the response, while the handover supplies the task-specific working context. The new chat can continue from the approved state without treating discarded ideas as current instructions.
Create and review the handover manually first
For a fuller manual workflow, use the AI conversation summarizer guide. If the original ChatGPT thread is still available, follow the fresh-chat continuation workflow.
- Write the objective in one sentence.
- List the active constraints and instructions that still matter.
- Record decisions already made, plus rejected options only when they prevent backtracking.
- Describe completed work and include the latest approved output or its essential details.
- End with unresolved issues and the exact next action.
- Review the result and remove stale, superseded, sensitive, or irrelevant information before pasting it into a fresh chat.
How thredly can assist—and where it stops
thredly creates a structured handover from conversation text supplied to it. The Chrome extension directly supports ChatGPT, Claude, and DeepSeek. The web app can create a handover from text you provide, and you can copy that result into another text-based AI conversation.
thredly does not replace Custom Instructions, provide persistent memory, or automatically sync context between chats or platforms. A handover is a compressed working document, so review it before use. If a conversation is simply becoming unwieldy, the provider-neutral conversation too long guide explains when staying put or starting fresh makes more sense.
Which option should you choose?
- Choose Custom Instructions when the guidance should remain broadly useful across conversations.
- Choose a thread-specific handover when the next chat needs the working state of one particular conversation.
- Use both when you need stable response preferences and precise task continuity.
- Review either setup when your preferences or project state changes; stale guidance can be worse than concise, current context.
Related guides
Frequently asked questions
Do Custom Instructions remember previous conversations?
Custom Instructions can guide responses across chats when enabled, but they do not themselves reconstruct the decisions, latest output, current progress, and unresolved actions from a previous conversation.
Can Custom Instructions replace a conversation summary?
Usually not for a task-specific continuation. Custom Instructions suit standing preferences, while a conversation summary or handover carries the current working context from one chosen thread.
Should I use Custom Instructions and a handover together?
Yes, when both are useful. Keep stable preferences in Custom Instructions and paste a reviewed handover into the fresh chat for the specific task you want to continue.
Do Custom Instructions carry over decisions from an old chat?
Only if you manually add those decisions to the standing instructions, which can make them stale or affect unrelated chats. A thread-specific handover is usually the clearer place for project decisions and current progress.