AI memory explained: what ChatGPT, Claude and Gemini remember

ChatGPT, Claude and Gemini can all remember things about you between chats, but each one does it differently, and none of them remembers your team. AI memory is personal: it lives in one account, it is built from one person's chats, and it does not reach a colleague, another tool or an agent.

This guide covers what each assistant stores, how to control it, and what that means when you use AI for work. Product details are taken from each vendor's own help docs as of October 2026. These features change often, so check the linked pages before you set policy.

What "AI memory" actually means

Four different things get called memory:

• Instructions you write. Standing guidance such as "write in US English" or "I run marketing at a B2B software company." The assistant applies it to every chat.

• Memories the assistant saves. Facts it picks up from your chats, or that you ask it to keep.

• Past-chat recall. The assistant looks back at earlier conversations when they seem relevant.

• Projects. Workspaces with their own files, instructions and chat history, kept apart from your other chats.

None of these is the same as the context window. Inside a single chat, the model sees the whole conversation. When the chat ends, only what was saved carries over. That gap is why every AI chat starts from zero unless something fills it.

ChatGPT

OpenAI's Memory FAQ describes two parts:

• Saved memories. Details you ask ChatGPT to remember, or that it saves as useful context. They are stored separately from chat history, so deleting a chat does not delete a saved memory that came from it.

• Reference chat history. When on, ChatGPT uses relevant information from past chats. What it draws from history can change over time as ChatGPT updates what it treats as useful.

You manage both under Settings, then Personalization, then Memory. When ChatGPT uses personal context, a Sources list may appear below the answer, showing which past chat, saved memory, custom instruction or file it drew on.

Temporary chats do not create or update memories. In Business and Enterprise workspaces, OpenAI describes a project-only memory option: chats inside a project can reference each other, but not chats outside it. Enterprise owners can turn memory off for the whole organization.

Claude

Anthropic's help article on chat search and memory describes two features:

• Chat search. You ask Claude to find something from a past conversation ("What did we decide about the launch date?"). This is on Pro, Max, Team and Enterprise plans.

• Memory. Claude saves memory as individual topics while you chat, rather than summarizing a chat after it ends. You can also say "remember this" to save something directly.

Memory is on by default for Free, Pro and Max. On Team and Enterprise, an owner enables it in organization settings, and members then opt in. You can view, edit and delete memory topics under Settings, then Memory.

Each project has its own memory space and project summary, separate from other projects and from non-project chats. Incognito chats are not saved to history or memory. On Team and Enterprise, memory entries appear in data exports and follow the organization's retention policy.

Gemini

Google's help page on memory from past chats describes a Memory setting under Personal Intelligence. When it is on, Gemini learns from your chats and may bring up earlier topics. It has clear limits:

• It needs a personal Google Account. It is not available on work, school or supervised accounts.

• You must be 18 or over, and Keep Activity must be on.

• It is not available in some features, such as Gems and Live chats.

Separately, Instructions for Gemini (called Saved info in some regions) lets you add instructions that apply to every chat. If you delete a chat from Gemini Apps Activity, Google notes there may be a short delay before Gemini stops using it. Google also offers temporary chats that are not used for personalization.

For a work account, the practical point is simple: Gemini's chat memory, as documented today, does not apply.

Side by side

What memory does not do for a team

Memory makes one person's assistant more useful. It does nothing for the team. Four gaps show up fast:

1. It belongs to one person. In March you decide to drop the monthly plan and sell annual only. You tell your ChatGPT. In May a colleague asks Claude to draft a pricing FAQ. Claude has no idea the decision exists.

2. It stays in one tool. What Claude remembers does not follow you into ChatGPT, Cursor or a custom agent. Anthropic lists an experimental import feature for moving memory between services, which helps one person switch tools. It does not share anything with a team.

3. It is inferred, not reviewed. The model decides what to keep. There is no owner, no date and no link to the document or meeting where a fact came from. Stale memories stay until someone notices.

4. It has no team permissions. Personal memory has no way to say "finance can see this, contractors cannot."

For the technical options behind this, see agent memory vs RAG vs fine-tuning.

How to use AI memory well at work

1. Review saved memories once a month. Delete anything stale or wrong, such as an old job title or a pricing plan you retired.

2. Use projects to separate work. One project per client or initiative keeps context from leaking between them.

3. Use temporary or incognito chats for one-off or sensitive tasks you do not want remembered.

4. Keep team facts somewhere shared. Decisions, pricing, positioning and style rules belong in a place with an owner and a date, not only in one person's assistant.

5. If you are an admin, check workspace settings. Know whether memory is on, who can turn it on, and how it fits your retention policy.

FAQ

Does ChatGPT memory carry over to Claude or Gemini?

No. Each assistant keeps its own memory inside its own account. Nothing syncs between vendors automatically.

Does deleting a chat delete what the AI remembers?

Not always. OpenAI says a saved memory is stored separately, so deleting the original chat does not remove it. Google notes a short delay before Gemini stops using a deleted chat. In Claude, you edit or delete memory topics directly in Settings.

Can my company see my AI memory?

It depends on the plan and the vendor. Anthropic states that on Team and Enterprise, memory entries appear in data exports and follow the organization's retention policy. For other tools, check your admin's settings and the vendor's business documentation.

Is AI memory the same as a bigger context window?

No. The context window is what the model can see in one conversation. Memory is what gets saved and brought into later conversations.

The short version

AI memory is useful, but it is personal, tool-specific and unreviewed. It helps you avoid repeating yourself to one assistant. It does not help your team share decisions across people, tools and agents. Treat personal memory as a convenience, and keep the facts your team depends on somewhere everyone, and every tool, can reach.

We're building Shane to keep team decisions and context in one place, with sources and permissions, and hand them to every AI tool your team uses. If that's a problem your team has, you can join the waitlist at tryshane.ai.