Does AI remember me?

You ask a chatbot for a dinner idea, and it starts with "Since you're vegetarian." You mentioned that once, weeks ago, in a different chat. It might feel helpful. It might feel a little startling. Maybe both. Either way, something saved that detail. So what, exactly, is doing the remembering?

The artificial intelligence (AI) model itself doesn't keep a diary about you. What it knows was set during training, before you ever typed a word, and your chat doesn't change it as you go. Each time it answers, it works only from the text it's handed at that moment. Once the reply is written, the model keeps nothing of your conversation. Not even your name. The app is what stores your chat history.²

So how does it follow a conversation? Inside a single chat, the app sends the conversation so far along with each new message. Anthropic's documentation describes every turn as carrying all the previous conversation history, plus your newest message. That's how it can refer back to something you said ten minutes ago. It's rereading, every time, inside its context window: the amount of text it can work with at once.²

Remembering you from one chat to the next takes something extra. It's an app feature, usually called memory. ChatGPT is one example. OpenAI's help page says it can keep saved memories, which are details you've told it, like being vegetarian, and it can also draw on your past chats. Anthropic's Claude has its own version, with settings to view and edit what it has saved, and an incognito chat that doesn't save to memory. The details differ from app to app and from plan to plan, so the one you use may work a little differently again.¹³

The model doesn't soak those notes up. In broad terms, when you start a new chat, the app adds the saved details that seem relevant to what the model is handed, along with your message. To the model, it's just more text to read. That's how "since you're vegetarian" turned up in a brand new conversation. It was a note, passed along.¹²

You're in charge of that memory. In ChatGPT, as of October 2026, OpenAI's help page says you can do four things.

• See what's saved, or ask it what it remembers about you.
• Delete a single memory, or clear them all.
• Turn memory off completely.
• Use a temporary chat, which doesn't use or add to memory.

To remove something fully, OpenAI suggests deleting the saved memory and also the chat where you first mentioned it.¹

If you use a chatbot with memory, open its memory settings once in a while and read what it has saved about you. It takes a couple of minutes, and what's there may surprise you. Keep what helps. Delete what doesn't. And ask yourself one question as you scroll: is there anything here you'd rather it didn't know?