Why We Start with Settings
Welcome back! In Level 1 you picked an assistant. Before we start asking it things, we're going to check its settings.
Why now? Picture moving into a new flat. Before you unpack, you'd check the locks and who has a spare key. Your assistant's settings work the same way. They decide who can see what you type and how long it's kept.
This takes about ten minutes, and you only need to do it once. Open your assistant and follow along.
One quick note first. Apps get updated all the time, so a menu name might be a little different on your screen. If you can't find something, look for Settings and then for words like Privacy, Data or Memory. They're nearly always nearby.
Step 1: Decide Whether Your Chats Help Train the AI
Here's something many people don't know. With most free assistants, your conversations can be used to improve the company's AI models, and this is often switched on unless you turn it off.
Is that bad? Not really. It's a choice. Some people are happy to help improve the tool, and others would rather keep their chats out of it. Either way, you should be the one who decides.
Here's where the switch lives in a few popular assistants:
- ChatGPT: click your profile icon, then Settings, then Data Controls, and look for "Improve the model for everyone."
- Gemini: find the setting called Keep Activity (it used to be called "Gemini Apps Activity"). When it's off, your future chats aren't used to train Google's models and don't show up in your activity history, although Google still keeps them for up to 72 hours to run the service.
- Claude: go to your Privacy settings and find "Help improve Claude."
- Copilot: go to your profile, then Privacy, where there are separate switches for training on your text conversations and on your voice conversations.
One thing to keep in mind: turning training off doesn't mean your chats disappear. In ChatGPT, for example, your chat history stays right where it is. The switch only controls whether your chats are used for training.
Step 2: Meet Your Assistant's Memory
Memory lets your assistant remember things about you from one chat to the next. Say you mention you're vegetarian. Next week, when you ask for a recipe, it might leave out the meat without you having to ask.
That's handy! But think of memory as a notebook the assistant keeps about you. You'd want to read that notebook now and then, wouldn't you?
Here's how to read it:
- ChatGPT: go to Settings, then Personalization, then Manage memories. You can also just say "forget that I…" in a chat.
- Claude: go to Settings, then Capabilities, then "View and edit memory." Memory from your chat history is on for everyone by default, but you can turn it off, pause it or reset it.
- Gemini: go to Settings, then Personal context, then "Your past chats with Gemini."
- Copilot: open Memory, then View memory, and click the bin icon next to any entry you want gone.
Step 3: Try It: Check and Delete a Memory
Let's actually do this. Open your memory list and read through it. If it's empty because you're brand new, that's fine. Here's a little test.
Start a normal chat and type this:
Please remember that my favourite colour is green.
Your assistant will probably confirm it's saved that. Now go back to the memory page. Can you see the new entry? Great. Now delete it.
How do you know it worked? Start a fresh chat and ask, "What's my favourite colour?" A good result is that it doesn't know any more.
There's a small catch. Some assistants also learn from your past chats in general, not just from the saved list. If it still seems to know, delete the chat where you said it too. We'll cover how to do that in a moment.
Want a variation? Try turning memory off completely for a day and see whether you miss it. There's no right answer here, only your preference.
Step 4: Delete a Single Chat
Sometimes you'll want one conversation gone. Maybe you typed something you regret, or it's just clutter.
In most assistants, your past chats are listed in a sidebar. Hover over a chat (or press and hold it on your phone), look for the three-dot menu, and choose Delete.
Copilot, for instance, lets you delete single conversations or your whole history whenever you like. Its history is otherwise kept for 18 months by default.
Step 5: Find the "Private Browsing" Option
You know how your web browser has a private or incognito window? Many assistants now offer something similar.
- ChatGPT calls it Temporary Chat. It isn't saved to your history, doesn't update memory and isn't used for training.
- Gemini also calls it Temporary Chat. These chats stay out of your activity and aren't used for training, but Google keeps them for up to 72 hours.
- Claude calls it incognito chat. Look for the ghost icon when you start a new chat. These chats are kept out of your history and memory and aren't used for training.
So when would you use one? A temporary chat is a good fit for a one-off question you don't want shaping your memory or sitting in your history.
One honest caveat, though. "Temporary" doesn't mean nobody ever sees it. Companies may keep these chats for a short time for safety reasons. So the best rule is still to leave really sensitive details out of any chat.
Key Takeaways
- Look for a training switch, such as "Improve the model for everyone," Keep Activity or "Help improve Claude," and set it the way you want. It's often on by default.
- Turning training off doesn't erase your chat history. Those are separate controls.
- Memory is a notebook your assistant keeps about you. Read it, delete entries you don't want, or switch it off.
- You can delete single chats from the sidebar menu.
- Temporary or incognito chats keep a conversation out of your history and memory, but they may still be kept briefly. Sensitive details are best left out completely.