What Is AI Chatbot Memory — and How Do You Control It?

AI chatbot memory is a feature that lets an assistant like ChatGPT, Claude, or Gemini carry facts about you — your job, your preferences, your ongoing projects — from one conversation into the next, instead of starting from zero every time you open a new chat.

What memory actually stores

Memory does not save your entire chat transcripts as a searchable diary, though some assistants keep a separate activity log that can also feed personalization. What it stores is closer to a running summary: short, distilled facts an assistant has picked up along the way, such as “works in accounting,” “prefers short answers,” or “is planning a trip to Batumi in September.” That summary is what gets quietly added to the background of your next conversation, so you don’t have to repeat yourself.

The three major assistants implement this a little differently. ChatGPT splits memory into two parts: an explicit list of “saved memories” you can read and edit one by one, plus a separate, non-listed ability to reference patterns from your past chats. Claude builds an account-scoped memory summary that updates automatically after each conversation, and lets you update it directly by telling it what to remember or forget — Anthropic rolled this out to Team and Enterprise plans in September 2025 and extended it to Pro and Max users the following month. Gemini stores similar personal context under what Google now calls Personal Intelligence, drawing on your Gemini Apps Activity within your Google Account, and requires that activity setting to be switched on to work at all.

Why it exists

Without memory, every new chat is a blank page — you re-explain your role, your writing style, or the project you’re working on each time. Memory is meant to remove that friction: the assistant adjusts its answers based on what it already knows about you, the same way a colleague who has worked with you for months needs less context than someone new. For anyone who uses a chatbot daily rather than for one-off questions, that adds up.

How to see, edit, and delete what it remembers

All three assistants put memory controls in account settings rather than burying them in a support ticket. In ChatGPT, it’s under Settings → Personalization → Memory, where you can view the saved-memories list, edit individual entries, or clear everything at once. In Claude, it’s Settings → Capabilities → Memory, showing a single summary you can edit conversationally or turn off per project. In Gemini, personal context sits under Personal Intelligence in your Google Account settings, alongside the separate Gemini Apps Activity log.

Turning memory off is not the same as deleting your chat history — the two are stored separately in all three products, so switching memory off stops new facts from being added without erasing past conversations. If you want a single conversation to leave no trace at all, use the temporary or incognito mode each assistant offers: Claude’s Incognito chats, ChatGPT’s temporary chat, and Gemini’s Temporary Chat all skip both memory and history for that session.

What to watch for

Memory is private to your own account — none of the three companies shares it with other users, and none uses it to identify you to third parties. But it does mean the assistant is building a standing profile of you over time, which is worth treating with the same care as any other place you store personal information: review what’s been saved occasionally, and clear anything you would not want summarized back to you, or to a support agent troubleshooting your account, later. It is also a newer, less battle-tested feature than the chat itself, so treat any single “fact” it recalls as a starting point to verify rather than as settled truth, the same caution worth applying to any AI hallucination. For a broader look at how these companies handle what you type into them, see what happens to your data when you use AI tools.

In the news

Memory sits alongside other ways AI companies are exposing more of what an assistant knows and does — Anthropic’s Claude Reflect usage dashboard, for instance, shows a similar shift toward giving users visibility into how much an assistant is tracking about their work.