Yes, both ChatGPT and Claude can recall facts about you across separate conversations by default, through two different systems. Docognito is not built for this. It addresses what someone glancing at your screen sees right now, not what either assistant recalls later. Review, edit or pause memory inside each assistant's own settings.
How ChatGPT remembers things about you
ChatGPT keeps two related but distinct systems running by default. The first is saved memories: discrete facts the assistant writes down as you chat, such as your role, a project you mentioned, or a preference you stated once. The second is chat history referencing, which lets the assistant draw on the content of your earlier conversations even when nothing was explicitly saved as a memory.
The practical effect is that a fact you mentioned weeks ago in an unrelated chat can shape an answer today, without you asking for it. This is presented as a convenience, and for many people it is one, but it means a new conversation is rarely a blank page.
How Claude's memory differs
Claude's approach arrived later. Anthropic introduced automatic memory entries in March 2026, and where enabled, Claude now writes its own record of relevant details as a conversation happens, without a separate save action from the user. Those entries can then be referenced in a future, unrelated chat the same way ChatGPT's memories can.
The mechanism is different from ChatGPT's split between saved memories and history referencing, but the outcome a user experiences is similar: open a new conversation, and the assistant may already know something about you that you did not type into that particular chat.
What people get wrong about how it works
The assumption that trips up most new users is that a fresh conversation starts from nothing. It rarely does, once memory is switched on by default.
| What people assume | What actually happens |
|---|---|
| A new chat starts with no context about me | Saved memories or memory entries can already be attached the moment the chat opens |
| Deleting a conversation removes any memory it created | A memory entry, once written, generally persists on its own, independent of the chat that produced it |
| Memory only stores what I explicitly ask it to save | Both assistants can write memory entries automatically, from ordinary conversation, without a separate save step |
| Temporary chat mode is already the default | It has to be selected each time, before you start typing, or the normal memory behaviour applies |
The second row is the one that catches people out most, since it means clearing your chat list is not a way to undo what memory already learned from it. Removing a memory entry has to happen through the memory settings directly, not through the chat history.
Where to review or turn off memory
Both companies put the controls in account settings rather than inside the chat itself, which is why most people never see them until they go looking.
Open ChatGPT's Data Controls
Under Settings, then Data Controls, you can view saved memories individually, delete specific ones, or turn the memory feature off entirely. Chat history referencing has its own separate toggle in the same panel.
Open Claude's memory settings
Claude's memory controls sit in Settings under the Privacy or Memory section, depending on your account. You can view what has been recorded, remove individual entries, or disable the feature for future conversations.
Use a temporary chat when you want a clean slate
Both assistants offer a temporary or incognito conversation mode that is excluded from memory and from your regular chat history, selected before you start typing rather than after the fact.
What Docognito does not touch
the conversation itself, which OpenAI and Anthropic still receive and store sits entirely outside what a browser extension can reach, and memory is a further layer on top of that record, generated and stored by OpenAI or Anthropic on their own servers. Docognito cannot read a memory entry, cannot clear one, and cannot pause the feature that creates them. There is no button in the extension that reaches into either company's account settings, because the extension has no access to them.
Turning off memory happens inside the assistant, not inside the extension
If your goal is stopping either service from carrying facts about you into new conversations, that setting lives in ChatGPT's Data Controls or Claude's account settings. Docognito changes what the page looks like while a tab is open, nothing about what either company stores.
Why this is a different problem than a glance
Memory is about what an assistant knows about you across time. The problem Docognito addresses is narrower and more immediate: what a person walking past your desk registers when their eyes cross your screen for a second. A document-shaped page answers that question. It says nothing about whether ChatGPT remembers your job title from a conversation three weeks ago, and it was never meant to.
People sometimes assume that because a tool hides the interface, it must also be quietly managing what gets stored behind it. That assumption is worth correcting early, since acting on it would leave the actual memory setting untouched while giving a false sense that it had been handled.
Reviewing memory versus changing what a screen shows
Knowing what either assistant recalls about you, or making that recall stop, means going into ChatGPT's Data Controls or Claude's account settings and reviewing the memory entries directly. If your concern is what a person can see on your screen in the moment, that is what Docognito is for. Treat the two as separate tasks, because the systems behind them are separate too.