Summaries, preferences, or other information
Select based on current task
Only then can this model reference it
Memory: Information an app may save across conversations and provide again; the scope and usage depend on the product.
Across a New Conversation, What Information Persists
The app may remember "you are a teacher" or a long-term preference, then provide it again in future conversations. It may also only save the current chat history, or have cross-conversation memory disabled.
Even if it saves your profession, it doesn't necessarily know today's lesson topic, students' level, or class duration. New task conditions should be restated; don't assume that "knowing who I am" means "knowing what I need this time."
How It Works with Context
Context is the information the model can reference this time. Memory is stored on the app side; the app needs to retrieve relevant content, then decide whether to use it for this input.
You can check two things separately: what was saved? and what was actually used this time? Saved content may be outdated, or it may be irrelevant to the new task.
When Using It
Check your own application's documentation for memory or preference features to confirm if they're enabled, what's saved, and how to remove data. Significant new tasks should always include current conditions—for example, if the grade level has changed, you shouldn't reuse old class information.
If results keep using incorrect background, first fix the current request, then check whether the application has stored information that needs updating. Specific options vary by product; this article doesn't provide unverified steps for universal buttons.
Keep in Mind — Memory Doesn't Mean Perfect Recall
The application can save summaries, preferences, or retrieval indexes; it doesn't necessarily keep every original text in the model. Storage duration, retrieval methods, and deletion effects depend on the specific system.
Memory also differs from model retraining. It's a way applications manage information, so you shouldn't conclude the model has permanently mastered all personal data.
Next Steps
Return to Context to understand how information enters the current task.
Reflect
If an application saves "I am a teacher," does the next lesson preparation automatically know the grade level and duration?
Reference answer: Not necessarily; professional information alone can't determine the current teaching task, so specific conditions still need to be provided.
Sources and Scope
Based on OpenAI · Conversation state. Technical details verified on 2026-09-09; examples and teaching organization are original content. Verify specific product support and operations separately; this article makes no cross-product feature guarantees.