News · Gemini Advanced adds cross-conversation recall with per-session source attribution
Gemini Advanced adds cross-conversation recall with per-session source attribution
Google's memory feature turns past chats into retrievable context — and puts the retrieval surface, not just the model, into the frontend.
What Gemini now does when you return to it
The change is narrow and concrete: Gemini can recall past chats to inform a response. Google frames two specific behaviors — answering a question about something you already discussed, and summarizing a previous conversation on request. The stated goal is removing two frontend frictions: starting over from scratch, and manually hunting for an earlier thread.
A third behavior is subtler and more useful for anyone doing sustained work: building on top of previous conversations or projects. That reframes each session from an isolated transcript into something closer to a continuous workspace. For frontend teams, this is the difference between a chat log and a stateful document.
Memory that announces itself in the UI
The most engineering-relevant line in the post is about attribution: 'Gemini may indicate when it uses your past chats in sources and related content.' Rather than silently blending prior context into a response, Google is surfacing provenance through the same sources-and-related-content affordances it already uses for external references.
Gemini may indicate when it uses your past chats in sources and related content.Montana Labs
That design decision matters because recall without visibility is a trust problem. When a model suddenly 'knows' something the user didn't restate in the current session, the interface needs to explain where that came from. Treating a past chat as a citable source — the same UI slot as a web reference — is a defensible way to make memory legible instead of spooky. The hedge 'may indicate,' though, means attribution isn't guaranteed on every recall, which leaves a gap between what the model uses and what the user sees.
Storage controls sit next to the feature, not buried
Google pairs the recall capability with an explicit control layer: users can review, delete, or set how long chat history is kept, and can turn off Gemini Apps Activity entirely through My Activity. The framing — 'You're in control over what information is stored' — ties memory directly to a retention setting rather than making it an invisible always-on behavior.
For anyone building similar features, the pattern is worth noting: the thing being remembered (chat history) and the retention policy governing it are the same object. Deleting the history removes the recall substrate. There's no separate 'memory' abstraction to reason about — retention and recall are the same lever.
A staged rollout that signals where this is really aimed
The feature launches in English, on web and mobile, for Gemini Advanced subscribers on the Google One AI Premium plan. Google says it will expand to more languages and to Workspace Business and Enterprise customers over the coming weeks. The sequencing is telling: consumer premium first, then the organizational tiers where cross-conversation memory has the most compounding value.
The specific implication for teams shipping conversational products: recall is becoming a baseline expectation, and the differentiator is the frontend contract around it. Google's bet is that users will accept persistent memory only if the interface both shows its work — through source attribution — and hands over deletion and retention controls in the same breath. That contract, not the recall itself, is the part worth copying.
Find this story relevant to you?
Contact us to find a unique solution
Need an AI engineering partner that can actually build?
We help businesses integrate AI, build AI-powered products, automate high-value workflows, and modernize the software systems behind them.
Related reading
More analysis around product delivery, operational AI, and the systems work that makes deployment hold up in reality.