FYI, context is the information supplied for this response.
Context is the information available to a model when it generates a response, including instructions, messages, and selected tool results or documents. A product can carry information across turns, but stored history is not necessarily included in every model call.
Context is what the application supplies for this response. Stored history or preferences become context only when the application includes or retrieves them.
View diagramSee it happen
Same request. Different information.
A model can use what is supplied for this response. Try adding a preference you mentioned earlier.
“Suggest something for dinner.”
Earlier preference: not included
Any dietary preferences I should take into account?
Context is not the same as permanent memory. An app may save preferences and include them later, but that is a separate feature.
Scripted illustration, not a live AI result. Example details are fictional.
Same request, different information
Imagine asking for a restaurant recommendation. Now add: “Under $40 per person, walking distance, and quiet enough for a conversation.” Those details change what a useful answer looks like.
They become part of the context if the application includes them with the model request. So can previous messages, a saved preference, a menu, or the result of a search. A model cannot reliably use a private preference that the application never supplies or makes accessible through a tool.
This is a fictional example, not a test of a particular product.
A common mix-up
Context and memory are not interchangeable. Memory usually means information a product stores for later use. Context is what is supplied for the current response. The product may retrieve a stored preference and put it into context.
That does not mean every system forgets everything after each turn. Applications can retain messages, retrieve older material, or carry a summary forward. Equally, seeing a message in your chat history does not guarantee that its full text reaches every later model call.
Why it matters
Give the relevant constraints, not just more text. A short, current brief can be more useful than a large pile of unrelated documents.
For the restaurant request, check that the budget and location are still included. If a conversation changes direction, restating the important constraints is often more useful than repeating the entire discussion.