Google Gemini Lets Users Review and Modify Stored Personal Data
The AI assistant's new memory-summary tool offers direct access to the profile it builds from user interactions, with controls to edit or delete individual facts.
A Window Into the AI's Profile
Google Gemini now offers users in the United States a direct view of what the assistant has learned about them. The memory-summary feature, available to paid Google AI subscribers, generates a categorised list of personal details the system has compiled from past interactions.
The tool surfaces information ranging from dietary restrictions to career details and pet names. Users can rewrite or remove any line in the summary, similar to editing a document, or issue text prompts to add or subtract facts. Changes are saved to a memory history with timestamps.
The feature addresses a tension inherent in personalised AI: the more context an assistant retains, the more useful its responses become, yet that same depth of knowledge raises questions about control and visibility. Gemini's memory system operates in two layers, and understanding both is necessary to manage what the assistant knows.
Two Tiers of Retention
The first layer, called Keep Activity, determines whether conversations with Gemini are saved and linked to a Google Account. For users aged 18 or older on personal accounts, the toggle is enabled by default. When active, past chats remain accessible across devices and can be resumed at any time. Disabling Keep Activity causes conversations to disappear after 72 hours and prevents them from following users between devices.
The second layer, which Google labels Memory, allows Gemini to construct a profile from the content of saved chats. This profile informs how the assistant tailors its responses. If a user mentions lactose intolerance, for instance, Gemini will avoid suggesting milk-based products in future exchanges. Memory cannot function unless Keep Activity is enabled, since the system requires access to past inputs to build its understanding.
Editing What Gemini Knows
To access the memory summary, users navigate to the settings menu in the Gemini web application, select Personal Intelligence, and then click Memory summary. The assistant generates a bulleted breakdown of major details it has recorded, organised into categories such as personal data, career information and interests.
From this view, clicking the edit button in the top-right corner allows users to modify any line directly. Alternatively, a text box at the bottom of the window accepts prompts for additions or deletions. Once changes are finalised, clicking Save commits them to Gemini's memory. The system retains a history of these edits, including timestamps, so alterations are tracked even as the profile evolves.
Turning Off Memory or Deleting Chats
Users who prefer not to have Gemini build a profile can disable the Memory toggle in the Personal Intelligence menu. This stops the assistant from committing new facts, though it does not erase information already stored. Existing details can be edited through the memory summary or removed by deleting the chats from which they were drawn.
To delete conversations, users open the Activity menu from settings and scroll through interactions by date. Clicking the "X" next to any chat removes it from the pool of context Gemini uses for personalisation. Combining these options allows users to shape both what the assistant remembers and how it applies that knowledge.
The Trade-Off Between Utility and Transparency
The memory-summary tool reflects a broader challenge for AI assistants: balancing the convenience of personalisation with the need for user oversight. Gemini's Personal Intelligence feature draws on Gmail, YouTube, Search and other Google services to enrich its responses, and the memory system adds another dimension by learning from direct conversation.
The more intimate the profile, the more relevant the assistant's suggestions become. Yet that intimacy also demands transparency. The ability to see and edit what Gemini has learned offers a measure of control, though it requires users to actively engage with the settings rather than relying on defaults.
For now, the memory-summary feature is limited to the United States and requires a paid Google AI plan. The rollout suggests Google is testing how users interact with visibility tools before expanding access. Whether the feature will extend to free-tier users or additional markets remains to be seen.
What This Means for AI Personalisation
The introduction of editable memory summaries marks a shift towards making AI personalisation less opaque. Rather than relying on users to infer what an assistant knows from its responses, the tool makes that knowledge explicit and actionable.
This approach may set a precedent for other AI platforms, particularly as regulators in multiple jurisdictions scrutinise how personal data is collected and used. Offering users the ability to review and modify stored information aligns with emerging expectations around data transparency, even as it places the burden of oversight on individuals.
The effectiveness of the tool will depend on how many users take advantage of it. Default settings tend to shape behaviour, and if most subscribers leave Memory enabled without reviewing the summary, the feature's impact on user control may be limited. Still, the option exists, and for those who choose to engage with it, Gemini's memory becomes less of a black box and more of a negotiable record.



