What is a context window?

In this guide, you will learn what a context window is in everyday terms, why long conversations seem to forget things, and how to work comfortably within the limit.

The desk analogy

Picture the AI working at a desk. Everything it needs to answer you, your question, the earlier parts of the conversation, and any documents you have pasted in, has to fit on that desk at once. The desk is generous, but it is not infinite.

This working space is called the context window. It is the amount of text the AI can see and use at any one moment. Anything on the desk, it can work with. Anything that has been pushed off the edge, it can no longer see.

The chatty AI tools you use, built on a large language model, all have a context window of some size. Newer ones have very large desks indeed, but the principle is the same for all of them: there is a limit, and it matters.

Why long chats seem to forget

Here is where the analogy earns its keep. As a conversation grows, more and more gets piled onto the desk. At some point the desk is full, and to make room for new text, the oldest text slides off the edge.

That is why a very long chat can suddenly seem forgetful. You mentioned your name, or a key detail, right at the start, and forty messages later the AI acts as though it never heard it. Nothing has broken. That early detail was pushed off the desk to fit everything newer.

Understanding this changes how you read that behaviour. The AI is not being careless or ignoring you. It genuinely cannot see the part of the conversation that has scrolled out of its window.

A quick word on how the limit is measured

You may see context windows described in tokens rather than words. A token is a small chunk of text, often a whole short word or a piece of a longer one. As a rough guide, 100 tokens is about 75 words, so a window of 128,000 tokens holds roughly 96,000 words, which is a fair-sized book.

You do not need to count tokens yourself. It is enough to know that the limit is real, that it is large but finite, and that long documents and long chats both eat into it.

Try it yourself

You can feel the edge of the context window with a small experiment. Start a fresh chat and give the AI a detail to hold on to:

Please remember that my project is called Bluebird and the deadline is
the 14th. I will refer back to these later.

Chat about other things for a while, the longer and more unrelated, the better, then come back and ask:

What was my project called, and when is the deadline?

In a short chat it will remember easily. The point is to notice that in a very long one, this is exactly the sort of early detail that can slip off the desk.

How to work with the limit

A common instinct, when a chat gets muddled, is to keep pushing on in the same conversation and hope the AI catches up. You will usually get better results by helping it out instead.

If a chat has grown long, start a fresh one and paste in a short summary of what matters. If you are working with a long document, ask the AI to summarise it in sections rather than swallowing it whole. Our guide on summarising a long document walks through that approach step by step.

Keeping the desk tidy, only the text that matters, is the single most useful habit for long or complex tasks. It costs a little effort up front and saves a lot of confusion later.

Next steps

Now that you understand the space the AI works in, a natural next step is to put it to use on real writing tasks. The next guide looks at using AI as a writing partner, where knowing about the context window helps you feed it the right amount at the right time.