Skip to content

What is context, and why does my chat get worse?

Context is everything the AI can see right now. It fills up, it gets crowded, and that is why a long chat starts giving worse answers. Here is the fix.

By Nick

2 min read

AI Basics Beginner
On this page

Before you start

Context is everything the AI can see right now: your messages, its replies, anything you pasted or uploaded, and the instructions behind the scenes. That is the whole of its world for this chat. It does not remember yesterday and it cannot peek at another tab.

Think of a desk. Every message puts another sheet of paper on it.

Why the desk fills up

The desk has a fixed size, and that size has a name: the context window. Early in a chat it is clear and the AI can see everything at once, so answers are sharp. An hour in, the desk is buried. Your first instructions are under forty sheets of back-and-forth, half of which were dead ends.

Two things happen then. It starts to miss things you said at the beginning, because they are buried. And it gets slower and vaguer, because it is reading the whole pile before every reply.

Pasting a huge document does the same thing in one go. One long PDF can fill most of the desk on its own, leaving no room to think.

What to do about it

Start new chats more often than feels natural. One job, one chat. When the job changes, the chat changes.

Carry over only what matters. Before you close a long chat, ask it to write the handover. The handover prompt is the one I use.

Put the permanent things somewhere permanent. Who you are, how you like answers written, what you are working on. That belongs in your settings or a project, not retyped into every chat. Tell it who you are covers that.

Try it on the chat you have open now

If a chat has gone stale, paste this before you give up on it.

This conversation has got long. In five bullet points, list the decisions we have made and
the facts I gave you that still matter. Leave out anything we abandoned. I am going to start
a fresh chat with just that list.
Example reply (trimmed). Your app will word it differently.

A long chat about planning a trip to Lisbon, plus the prompt above

  • Travelling 12 to 17 May, two adults.
  • Budget is $2,500 for everything except flights.
  • Staying in Alfama, flat booked at $140 a night.
  • Day trip to Sintra on the Wednesday, by train rather than a tour.
  • One of us is vegetarian, so look for places with real options.

Paste them into a new chat and carry on.

Common questions

Is context the same as memory?

No. Context is what is in front of it in this chat. Memory is a separate feature where the app keeps a few notes about you between chats. Context is the one that fills up.

How long is too long?

When answers start repeating themselves, ignoring something you said earlier, or getting vaguer, you are there. That is the signal, not a word count.

Does a new chat lose everything?

Yes, which is the point. Carry over what matters with a handover prompt and leave the rest behind.

What to do next

Related reading

Last reviewed 9 October 2026 by Nick.

Software changes fast. If something here is out of date, tell me and I'll fix it.

Get new guides by email

Occasional and practical. No spam, no hype, unsubscribe whenever.

Occasional practical guides from UncomplicateAI. You will get a confirmation email before anything else is sent. Unsubscribe any time. Privacy

You are not subscribed until you confirm the email. Unsubscribe any time.