B5 · Generative AI and Conversational Assistants

Chat History, Context and Memory

Context is everything a chat assistant can see while it answers: your current message plus the earlier messages in the same chat. A new chat usually starts with no context. Some assistants also have a memory feature that saves chosen details across chats; you can usually see, change or switch it off in the settings.

10 min Beginner Jump to the activity

What you need: Nothing. This page starts from the beginning, with no programming and no formulas.

Definition

You tell a chat assistant 'I'm vegetarian' and ask for a lunch idea, and it suggests paneer. Then you open a new chat and ask again. Will it still remember?

Context is everything a chat assistant can see while it answers: your current message plus the earlier messages in the same chat. A new chat usually starts with no context. Some assistants also have a memory feature that saves chosen details across chats; you can usually see, change or switch it off in the settings.

In simple words

A chat is like a conversation on one page of a notepad. While you keep writing on the same page, the assistant can read what you wrote above. Start a fresh page, and it usually can't see the old one, unless you have asked it to keep a note that it carries from page to page.

Why it matters

Knowing what an assistant remembers helps you get better answers and protect your privacy. You'll know when to repeat important details, why a very long chat can lose early details, and why you should never share private information you wouldn't want saved.

Key terms

Context
The information the assistant can use while answering, including earlier messages in the chat.
Chat history
The list of earlier messages in a conversation.
Context limit
The most text an assistant can take into account at once; in very long chats, early details may be missed.
New chat
A fresh conversation that usually starts without the earlier messages.
Memory feature
An optional setting in some assistants that saves selected details across chats, which you can usually view and delete.

Explained in detail

Inside one chat

When you send a message, the assistant does not read only that message. It also reads the earlier messages in the same chat, which is why you can say 'make it shorter' and it knows what 'it' means. If you said earlier that you are vegetarian, it can use that when you later ask for a lunch idea. Everything it can see while answering is called its context.

Starting a new chat

A new chat usually starts empty. The assistant cannot see what you said in another chat, so it may suggest a chicken sandwich even though you told it, elsewhere, that you are vegetarian. You need to tell it again. This is not forgetting the way a person forgets; the earlier messages are simply not part of the new chat's context.

Memory features and limits

Some assistants offer a memory feature that saves selected details, such as your preferred language, and uses them in new chats. Where it exists, the settings usually let you see what is saved, delete it or switch it off. Even within one chat there is a limit to how much text the assistant can consider at once, so in a very long chat it may lose track of something you said at the start. If a detail matters, repeat it.

SituationUsually remembered?What to do
Later in the same chatYesRefer back naturally
A new chatNoRepeat the important details
A new chat with memory switched onSome saved detailsCheck the settings to see what is saved
A very long chatEarly details may be missedRepeat key details near your question

Privacy

Because chats can be stored and memory can save details, share only what the task needs. Use made-up details when practising, and never type passwords, phone numbers, addresses or other people's private information.

Two habits make chats work better. Start a new chat for a new topic, so that old details do not confuse the answer. And put the details that matter in your first message: who the answer is for, what you already know and any limits, such as length. If an answer ignores something important, the simplest fix is to say it again.

How it works

  1. In one chat, give a made-up preference, such as 'I'm vegetarian and I like spicy food.'
  2. Ask a question that depends on it, such as 'Suggest a lunch.' Note whether it used the preference.
  3. Open a new chat and ask the same question without repeating the preference.
  4. Note what it remembered and what you had to repeat.
  5. If the tool has a memory setting, look at what it has saved, with an adult's help.

Worked example

In this example, chat 1 begins with 'I'm vegetarian and I love spicy food,' followed by 'Suggest a lunch for tomorrow.' The assistant suggests a spicy chana wrap, so it used both details. In chat 2, a new chat, the learner asks only 'Suggest a lunch for tomorrow.' The assistant suggests a chicken sandwich: neither detail was in the new chat's context. The learner records: 'Same chat: remembered both. New chat: remembered neither, so I had to repeat them.'

Try it yourself

Same chat, new chat

Follow one preference through two chats.

Will the assistant usually know this?

Sort each situation.

  • Something you said at the very start of an extremely long chat

  • You start a new chat and ask for a lunch idea

  • Your preferred language, saved in a memory setting you switched on

  • A detail you shared in a completely different app

  • Earlier in the same chat, you said you're vegetarian

  • You say 'make it shorter' right after it wrote a paragraph

A new case: planning a trip in two chats

Use made-up details only, with an adult's permission.

  1. In one chat, describe a made-up trip: the city, a small budget and that you dislike long walks.
  2. Ask for a one-day plan. Did it respect all three details?
  3. Start a new chat and ask for a one-day plan for the same city. Which details did you have to repeat?

Think about it

  • Write one rule for yourself about what to repeat when you start a new chat.

Common mistakes

  • Expecting a new chat to know what you said in another chat.
  • Assuming a very long chat still keeps every early detail in mind.
  • Sharing real private details because the assistant 'remembers' things.

Check yourself

Question 1

1. You open a new chat. Does the assistant usually know what you said in an old chat?

Question 2

2. Why does 'make it shorter' work in the same chat?

Question 3

3. What is safe to share when practising?

Question 4

4. New case: in a very long chat, the assistant ignores the word limit you set at the start. What should you do?

Summary

  • Within one chat, the assistant can use earlier messages: that is its context.
  • A new chat usually starts empty, unless a memory feature saved some details.
  • Repeat important details, and share only made-up or necessary information.

Finished reading?