NotebookLM Memory: What Carries Between Research Sessions
NotebookLM, now Gemini Notebook, keeps notebook material and retained chat history. Learn what that means for the next research session and what to save yourself.
Updated 6 min read
NotebookLM memory is a combination of saved notebook material and conversation history, rather than a promise that every notebook remembers everything you have ever researched. Google’s current chat help says history is retained and private to you. For the project result you want to carry into other work, fenbs can hold a manually saved summary, decisions and next steps outside the research conversation.
Google now calls the product Gemini Notebook (opens in a new tab). This guide uses both names because the useful question is the same: what will be available when you return, and what will another person or assistant actually know? A saved source, an earlier question and a checked project conclusion do different jobs.
What can carry over when you reopen a notebook?
Start by checking the notebook itself. Find the sources, notes and earlier conversation you expect to use. Then check which material is selected for the next question. Having a document in a collection does not establish that the answer used the relevant passage or that the document still describes the current situation.
| Record | What to check when you return | What it does not establish |
|---|---|---|
| Notebook sources | The intended source is present, available and selected. | Every future answer will include every relevant passage. |
| Retained conversation | You opened the intended notebook and can find the earlier exchange. | A colleague has read your private conversation. |
| Research notes | The note states its evidence and is included when needed. | An assistant’s interpretation has been approved by a person. |
| Project handoff | The summary names the result, uncertainty and next action. | The underlying research will remain current without review. |
For example, a saved supplier brochure can support tomorrow’s comparison. It does not record that your team rejected one option after a sample test unless someone also saves that result. Treat research materials as evidence and the project handoff as an account of what your team did with the evidence.
Is chat history saved between sessions?
Google’s current chat documentation (opens in a new tab) says chat history is retained and private to you, with a Delete Chat History control in the chat panel. It also documents Save to note for preserving a response with its formatting and clickable citations. Check the history visible in your own notebook before depending on a specific exchange.
Privacy matters for teamwork. A colleague should not need your private conversation to understand which conclusion is ready to use. Put the checked result in a record you intend to share, with the supporting source and any open question. Sharing background documents and communicating the current project position are separate tasks.
Avoid testing recall with a question that contains the answer. Instead of “Do you remember we rejected the aluminum option?”, ask, “Which option has been rejected, and what evidence supports that?” Read the cited material before treating the reply as a valid handoff.
Does saved material automatically become usable context?
Google’s FAQ on context (opens in a new tab) distinguishes sources, selected notes and conversation history in prompts. The practical lesson is to check selection and scope, rather than assuming everything you saved is being used in the same way. Ask a question specific enough to identify the passage or note that should support the answer.
The notes documentation (opens in a new tab) says newly created notes appear in the Sources panel and can be included or excluded with selection controls. It also explains how to convert older notes from the Studio panel into sources. If your important summary is saved but absent from an answer, inspect how that note is available to the conversation.
Keep your own interpretation visibly separate from quoted evidence. A note saying “this may meet our requirements” should not silently become “this meets our requirements” after several summaries. This is part of assistant memory: retained information still needs a clear meaning, scope and correction path.
How do you check that the sources are still current?
Open the document behind an important answer. Check its revision, whether you can still access it and whether the passage applies to your question. Google’s source guide (opens in a new tab) describes how supported Google Drive imports sync and how access changes can make a source unavailable. Other material, such as a pasted excerpt, can still reflect an older version of the original information.
- Choose the claim that affects the next decision. Do not spend the same effort rechecking harmless formatting preferences.
- Open the cited source and locate the supporting passage. Check whether it states the claim directly or requires an interpretation.
- Compare the source with the current original when freshness matters. Record any changed wording or missing information.
- Correct your research note and project summary together if the new evidence changes the result. Preserve what remains uncertain.
If a source disappears, do not repair the gap by asking the assistant to reconstruct it confidently. Find an available authoritative replacement or mark the conclusion as needing verification. Saved chat wording can tell you what was said; it is not a substitute for unavailable evidence.
What should you save at the end of a research session?
Save a summary that helps you resume without rereading the whole exchange. Say a three-person design studio is comparing materials for a retail display. The next session needs to know which specification was checked, which candidate remains possible and whether a person has approved ordering a sample. It does not need every exploratory question.
- Question: the specific decision the research supports.
- Evidence: the sources checked and what each one establishes.
- Result: the conclusion you can support now, labeled as tentative when appropriate.
- Approval: the choice a person actually made, separate from the assistant’s suggestion.
- Next action: the remaining check, who must answer and where to start next time.
Date a summary when its accuracy depends on changing information. Include enough source detail to revisit the claim, and remove irrelevant conversation fragments. fenbs keeps this deliberate work record in project context, tasks and Where we left off; its handoff workflow helps the next supported assistant identify what still needs doing.
Keep the research result ready for the next task
Manually save the checked conclusion and outstanding question in fenbs. Add the next action as a task and record where the project stands before stopping. A supported connected assistant can then read that handoff without needing access to your private notebook conversation.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
Can that record follow you to another assistant?
It can when you deliberately save and provide it. Do not assume a fresh notebook or another AI app has the same context just because the original conversation is retained. Ask the next assistant to restate the project’s goal, checked result and open question before it begins a dependent task.
Our NotebookLM alternatives guide covers choosing a research environment; that is a separate decision from preserving this session’s outcome. For portable AI memory, keep the project summary outside the private chat and link back to source material where appropriate. fenbs does not connect directly to Gemini Notebook or import its notebooks and chats.
How to do this with fenbs
- Create a free account, free for one person with no credit card, and add the project your research supports.
- Manually add a concise research summary to the project context. Include the source references, checked conclusion and unresolved question.
- Add the next research or delivery action as a task. Record a person’s approved choices in Decisions and rules, keeping unapproved suggestions separate.
- Connect a supported assistant through the integration guides. Ask it to read project context before working and save progress before it stops.
fenbs is not the right fit if you need a tool to analyze notebook documents or generate research outputs. Keep that work in your research tool. Use fenbs when the selected outcome needs to become an explicit next step for another session, and explore the AI memory topic for more ways context can persist.