Gemini’s editable memory gives a profile a separate life from chat history
A summary can be reviewed and corrected, but deleting it and removing past conversations remain different actions.
Gemini’s Memory Summary gives eligible users a place to inspect and correct the preferences, interests and ongoing goals the assistant uses to personalize responses. Google introduced the readable summary in its November 13 product announcement. The useful change is visibility: a mistaken assumption no longer has to be inferred solely from the answers it keeps influencing.
The feature is distinct from connecting another app to Gemini. A bank connection supplies external information; a memory summary describes what the assistant understands about its user. The controls answer different questions, even though both sit within Personal Intelligence.
A profile should be correctable
Google’s Memory Summary help page says users can add details, fix inaccuracies and remove information they no longer want saved. Updates are made with prompts, and Gemini may ask a follow-up question when a request is ambiguous. This gives the user an explicit way to change the assumptions behind personalized replies.
Consider a preference that was temporary: someone asked for vegetarian restaurant ideas while organizing a meal for a guest. If an assistant subsequently treats that as a permanent personal preference, its recommendations can become consistently inappropriate without any single answer appearing obviously broken. A readable summary makes that kind of misunderstanding easier to identify and correct. This is an illustrative use case, not a claim that Google demonstrated that particular error.
The distinction between a fact and a preference matters too. Speaking a language, planning a trip and preferring a short answer are different kinds of context, with different useful lifetimes. A correction surface is valuable because people change and conversations contain situational requests, not just enduring biographical information.
History makes changes inspectable
The documented Memory History view lists when specific information was added or modified. That can help a user connect a change in the assistant’s behavior to a change in its remembered context. The summary and its history serve complementary purposes: one shows the current understanding, the other helps explain how it developed.
The feature requires a personal account, an age of at least 18, Keep Activity enabled and Memory switched on. It is available in Gemini’s mobile and web apps where supported. Those conditions are more precise than simply saying that every Gemini account now has an editable profile.
Editing the summary is not deleting every record
The help page describes control over the contents of the memory summary. It does not say that removing a detail from that summary deletes every past chat, connected-app record or other copy of the information. Those are separate data-management questions, with separate settings and policies.
That boundary prevents a useful control from being oversold. Removing an outdated goal can improve future personalization without necessarily erasing the conversation in which the goal first appeared. Users need the distinction to understand which action they are taking, rather than assuming a single “remove” request operates across every layer of the account.
The most meaningful test of Memory Summary is consequently not how much it can remember. It is whether a person can recognize an inaccurate assumption, change it and understand the scope of that change. Personalization becomes more useful when the person being described has a readable way to disagree.