Chatbot Memory and Personalisation: What Does Your AI Remember About You?

Chatbots that remember

Early chatbots forgot you when the window closed. Many now keep a running picture of you: your job, projects, preferences and even family details. This memory makes replies feel tailored. It also means a chatbot account can hold a detailed profile of your life.

Google describes Gemini as able to personalise responses using three things: the memory of your past Gemini chats, your content and activity in certain Google apps you connect, and your preferences such as instructions on how you want it to respond. It says these features are not available to everyone and require a personal Google Account, not a work, school or supervised one.

How Claude’s memory works

Anthropic’s help page gives unusually specific detail. Claude builds memory as you chat, saving individual topics as they come up rather than summarising conversations after they end, and you can ask it directly to remember something. Everyday context such as your role, projects and communication preferences is kept. By default, sensitive subjects like health, politics and religion are excluded unless you switch on an “Include sensitive topics in memory” setting. Some information is never saved, including government ID numbers, financial account numbers and immigration status.

Other products work differently, so read each one’s own page. The pattern to watch for is the same: what the tool remembers is a choice made in settings, and the defaults may not match your comfort level.

The deletion trap

The most useful line on Anthropic’s page for privacy is this: deleting a conversation does not remove memories generated from it. You must delete those memories individually. People who clear their chat history and assume the slate is clean may be wrong. Memory is also included in data exports, which is handy for checking what is stored.

The controls it describes are worth knowing as a model for what to look for elsewhere:

• View and edit. You can open your memory in settings, then edit or delete any topic, or tell Claude in chat to remember, change or forget something.
• Pause. Stops memory being used or added to, but keeps what exists.
• Reset. Permanently deletes all memories, including project memories, and cannot be undone.
• Single chat. You can turn memory off before the first message of a chat.
• Incognito. Nothing from these chats is saved to memory or chat history.
• Projects. Each project has its own memory, separated from other projects.

Why it matters for privacy

A memory feature can combine details you gave on different days into one profile. If your account is taken over through a weak password or a phishing page, that profile goes with it. If you share a device, someone else may see answers shaped by your past. And if you told the tool something you later regret, it may still be stored.

OWASP’s entry on sensitive information disclosure tells users to avoid entering sensitive information and says consumers should know how to interact safely with LLMs. It also says users should be able to opt out of having their data included in training, and that data entered today may later appear in outputs, so inputs should be treated as potentially disclosable. Memory and training are separate settings, so check both. Our guides to AI chatbot privacy settings and data retention, what not to tell AI chatbots and how to opt out of AI training go step by step.

A sensible setup

Open the memory page and read it as a stranger would. Remove anything you would not want in a breach. Use incognito or no-memory modes for health, money and family issues. Protect the account with a strong unique password and two-step sign-in, because the account is now the key to your profile. If the account may have been exposed, change the password and follow the data breach checklist. Be careful too with AI companions, which are built around remembering personal details and which children may use.

This article reflects public information as of 2026-11-06. AI tools and rules change quickly; we update articles when new facts are published. Spot an error? See our editorial and corrections policy.