Does Gemini Notebook (NotebookLM) have memory? What it actually saves, and what it doesn't

What Gemini Notebook actually saves

Every message you send inside a given notebook is saved to that notebook's own chat history, and it's used as context for the answers that follow within that same conversation. Close the tab, come back a day later, and the conversation is still there. That sounds unremarkable now, but it wasn't always true: before the fix, which 9to5Google dated to 16 December 2025 ("NotebookLM rolls out chat history, adds AI Ultra tier"), refreshing the page or closing the tab could wipe the entire chat, a frustration well-documented enough that XDA Developers wrote it up as one of the product's biggest pain points. If you're reading an older complaint online about NotebookLM "forgetting everything" the moment you close the tab, that's describing a real bug from before the fix, not current behavior. Google's own March 2026 Workspace Updates post documents further customization built on top of that persisted history.

On 29 October 2025, Google shipped a chat upgrade alongside an expanded context window (up to 1 million tokens, an 8x increase) that TechRadar reported at the time as "6x longer conversation memory." Read that phrase carefully: it's describing how much of a single ongoing conversation the model can hold in context at once, a context-window sizing change, not a new feature that remembers you across separate notebooks or separate sessions with different source sets. The same update added a "Goals" system that lets you set a persistent response style or persona for a notebook, which is the closest thing to configurable personalization Gemini Notebook currently offers, but it's a setting you configure once per notebook, not something the model infers and remembers about you automatically.

What it doesn't do: no memory across notebooks

Each notebook is its own isolated environment. Facts, preferences, or context you've built up in one notebook stay in that notebook; opening a different notebook starts from zero, with no carryover of anything the model "learned" about you elsewhere. There's also no user-facing memory list anywhere in Gemini Notebook, nothing resembling ChatGPT's saved-memories panel or Claude's new topic-by-topic memory editor (Anthropic shipped that on 25 August 2026, letting users edit or delete what Claude remembers by topic rather than only as an all-or-nothing wipe). If you want Gemini Notebook to treat two notebooks as related, the product itself won't do that connecting for you.

The "Notebooks in Gemini" mix-up

This is where most of the genuine confusion comes from, and it's worth being precise about, because the two product names are almost identical. "Notebooks in Gemini," launched inside the main Gemini app (not Gemini Notebook itself) on 8 April 2026, is a project-workspace layer: dedicated spaces where you organize chats, custom instructions, and files around a topic. Google's own post lets sources sync bidirectionally between a Gemini-app notebook and a Gemini Notebook notebook, so a source you add in one shows up in the other, and you can start a task in Gemini and hand it off to Gemini Notebook's specialized tools (Audio Overviews, Video Overviews, Infographics) or vice versa.

Third-party write-ups, including pieces from ghacks.net and Medium contributors, describe this sync loosely as Gemini being able to "remember" what's in your notebooks. Google's own announcement post never uses that framing; it describes source syncing, not a memory system. The practical distinction matters: syncing means a document you saved is now visible to a different tool, on demand, when you reference it. Memory, in the ChatGPT or Claude sense, means the system proactively recalls a fact about you without you re-supplying the source. Gemini Notebook and "Notebooks in Gemini" together give you the first. Neither one gives you the second.

Side-by-side: memory across the major AI tools

Tool                        Memory scope                          User-editable?    Persists automatically?
--------------------------------------------------------------------------------------------------------
Gemini Notebook              Chat history, within one notebook     No                Yes, within that notebook only
"Notebooks in Gemini"        Source sync across the two products    No                Not memory, a sync feature
ChatGPT                      Saved facts, across all conversations Yes, a list       Yes, opt-out available
Claude (since 25 Aug 2026)   Saved facts, topic-by-topic editing    Yes, per topic    Yes, with a sensitive-topics setting

Sources: Google Workspace Updates (Mar 2026, Apr 2026), Google's Gemini Notebook
support FAQ, Anthropic's 25 Aug 2026 memory update, checked 27 August 2026.

The privacy angle nobody frames this way

Framed as a limitation, no cross-notebook memory sounds like Gemini Notebook is behind ChatGPT and Claude. Framed as a privacy property, it's arguably ahead of them for a specific use case. Anthropic's own reason for shipping a sensitive-topics setting and topic-by-topic editing on 25 August 2026 was giving users a way to exclude categories from memory retention and delete a specific memory without wiping everything, which is only a necessary feature in the first place because a system that remembers across every conversation needs a way to un-remember something. Gemini Notebook sidesteps that problem structurally instead of adding a setting to manage it: because Google's own support FAQ describes conversation history as an input to a single notebook's answers rather than a standing memory system, there is no cross-notebook channel for a fact to leak through to begin with. If you're specifically choosing a tool because you don't want AI memory carrying information between unrelated contexts, that's a genuine, if unadvertised, advantage of how Gemini Notebook is built, not an oversight Google needs to fix.

Why Gemini Notebook works this way

Gemini Notebook is built around a different premise than ChatGPT or Claude to begin with: it's source-grounded, meaning its answers are meant to come from documents you've explicitly added to a notebook, with citations pointing back to them, rather than from a general-purpose model drawing on whatever it recalls about you from past chats. Google's own Gemini Notebook support FAQ describes conversation history as one of several inputs used to generate an answer, alongside your sources, not as a standing memory profile of you as a user. A memory system that quietly injects facts from other notebooks would work against that grounding: if you're using Gemini Notebook for one client's research and another for a different client's, isolated notebooks are a feature, not a gap, since nothing about client A can leak into client B's answers via memory. The tradeoff is that if you genuinely want continuity, a running project spanning multiple notebooks, you have to build it yourself.

How to fake memory in Gemini Notebook, since it won't do it for you

• Save a summary as a Note, then convert it into a Source. Notes you write or generate inside a notebook can be promoted to a source, which means future chats in that same notebook will retrieve from it like any other document. This persists within the notebook, not across notebooks, but it's the closest built-in mechanism to "teaching" a notebook something durable.
• Use the Goals/customization setting (added 29 October 2025) to fix a response style or persona per notebook, so you're not re-explaining formatting preferences every session.
• Keep one running "memory palace" notebook as a single source of truth for a long project, and manually copy forward the key facts or decisions you want later notebooks to have access to. This is a manual workaround, not automation, but it's what heavy users on forums like XDA Developers report actually doing.
• Deliberately use the "Notebooks in Gemini" sync by adding a Gemini Notebook notebook as a source inside a Gemini-app notebook, so ordinary Gemini conversations can reference what's saved there. It's a manual bridge you set up once, not automatic recall.

If memory across conversations matters more than source grounding

If what you actually want is a single assistant that remembers your preferences and facts about you across every conversation without manual bridging, that's specifically what ChatGPT's memory and Claude's memory are built for, and our ChatGPT memory vs Claude memory comparison covers how each one actually works, what's editable, and where the privacy tradeoffs sit. Gemini Notebook is the better tool when the priority is answers grounded in documents you control, with citations you can check; it's the wrong tool if what you're after is a persistent personal memory layer, and no amount of workaround setup fully closes that gap.

People also ask

Why doesn't NotebookLM remember what I told it in a different notebook?

By design. Each notebook is an isolated environment with its own chat history; nothing carries over to a different notebook automatically. If you need continuity across notebooks, you have to copy the relevant information forward yourself, or use it as a source in the new notebook.

Does Gemini Notebook forget my chat if I close the tab?

No, not anymore. Conversation history is auto-saved and resumes when you reopen the notebook, a fix 9to5Google dated to 16 December 2025. Before that fix, refreshing or closing the tab could wipe the chat entirely, which is the source of most older "NotebookLM forgets everything" complaints you'll still find online.

Is "Notebooks in Gemini" the same thing as Gemini Notebook?

No, and the near-identical names are the whole source of confusion. Gemini Notebook is the renamed NotebookLM, a standalone source-grounded research tool. "Notebooks in Gemini" is a separate project-workspace feature inside the main Gemini app that syncs sources with Gemini Notebook. They're connected products, not the same product.

Can I see or edit what Gemini Notebook remembers about me?

There's nothing to see or edit in that sense. Gemini Notebook doesn't maintain a user-facing memory list the way ChatGPT and Claude both do; the only persistent thing per notebook is its own conversation history and any Goals/customization setting you've configured.

Does Gemini Notebook have anything like ChatGPT's memory or Claude's memory feature?

No. The closest analogues are per-notebook chat history persistence and the Goals customization setting, neither of which is a cross-conversation memory system. If cross-conversation memory is what you need, ChatGPT and Claude both offer it natively; Gemini Notebook does not.