Gemini Gems vs NotebookLM: Instructions or Source Research?
Gemini Gems give an assistant a reusable persona and instructions. Gemini Notebook (formerly NotebookLM) grounds answers in your uploaded sources. Here is which one fits which job.
Updated 4 min read
Gemini Gems and Gemini Notebook (formerly NotebookLM) are two different tools built around two different approaches to grounding an AI assistant. Gems give an assistant a set of reusable instructions and a defined role. Gemini Notebook ties answers directly to the specific documents you upload. Choosing between them depends on whether you need consistent behavior or grounded evidence.
Both tools operate inside Google's ecosystem, and neither one keeps track of the ongoing work across sessions on its own. A tool like fenbs holds the decisions, rules, and project progress that should stay available regardless of which surface you use.
What Gemini Gems do
A Gem is a saved version of Gemini with a specific name, set of instructions, and optional knowledge files. You create a Gem once, give it a persona and rules, and open it whenever you want that behavior without rewriting the system prompt. It is reusable across sessions.
Gems work well for recurring tasks: a writing assistant with your preferred style, a meeting-prep assistant with the right questions, a research assistant that follows a consistent format. The Gem carries the instructions; you bring the content for each session.
Gems are available in Google One AI Premium and through Google Workspace. Note that on personal accounts, Gems are scheduled to become skills from November 2026 — check Google's help center (opens in a new tab) for the current status before building a workflow around Gems.
What Gemini Notebook does
Gemini Notebook became the new name for NotebookLM in July 2026. It is a research tool. You upload documents — PDFs, Google Docs, web pages, audio files — and Gemini Notebook answers your questions using only those sources. Every answer is grounded in the material you provided, and it cites the relevant passages.
This makes it especially useful for reviewing contracts, studying a set of documents, understanding a report, or querying a collection of research papers. The assistant does not draw on its general training data for answers — it stays within your uploaded sources, which is its main strength and its main limitation.
Instructions vs. source grounding: the core difference
| Aspect | Gemini Gems | Gemini Notebook |
|---|---|---|
| Purpose | Reusable assistant persona and instructions | Grounded answers from your uploaded documents |
| What grounds the answers | Instructions you write, plus Gemini's general knowledge | Only the documents you upload |
| Best for | Recurring tasks needing consistent behavior | Analyzing or querying a specific set of sources |
| Citations | Not the focus | Yes, cites passages from uploaded documents |
| Reuse across sessions | Yes — open the Gem for any session | Per-notebook; sources stay in that notebook |
| Knowledge files | Optional; uploaded to the Gem | Required; the sources are the product |
A Gem answers with the knowledge Gemini was trained on, shaped by your instructions. Gemini Notebook answers only from what you gave it. If your question relies on something not in the uploaded documents, Gemini Notebook will say so rather than hallucinate a plausible-sounding answer.
Where knowledge files fit
You can add files to a Gem as well as to a Gemini Notebook, but they serve different roles. A Gem's knowledge file gives the assistant background to draw on alongside its general training — a style guide, a product catalog, a set of procedures. The assistant blends those with its general capabilities.
In Gemini Notebook, the uploaded sources are the boundary of what the assistant knows. It will not go outside them. This strictness is useful when accuracy against a specific document set is the requirement, and a creative answer would be worse than a "not found" answer.
Using Gems and Notebook together
The two tools are complementary. Use Gemini Notebook to query a set of documents and extract findings. Then use a Gem with your preferred writing or analysis persona to develop the findings into a report or presentation. The research phase and the production phase benefit from different grounding.
The gap is handoff. The conclusions you reached in Gemini Notebook exist only in that notebook's chat. Opening a Gem requires you to brief it on those conclusions manually, or copy them in as a file. For decisions and findings that matter across more than one session, keeping them in a shared record saves the re-explanation and prevents the two sessions from working from different versions of what was agreed.
For comparisons with other assistant customization tools, see Claude Projects vs Gemini Gems. For alternatives to Gemini Notebook, see NotebookLM alternatives.
Keep decisions from Notebook research available in Gem sessions
fenbs keeps the decisions you make from research sessions — the facts you verified, the conclusions you reached, the rules you set — in one place that connected assistants read before starting work. Whether your next session opens a Gem or a fresh Gemini chat, the conclusions from the Notebook phase are already there.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
How to do this with fenbs
- Create a free account and add your project with its goal and current tasks.
- After a Gemini Notebook session, record the key findings as task notes and the decisions as Decisions and rules in fenbs.
- Connect your AI apps over MCP from Integrations so any future assistant session reads the same project context.
- When you open a Gem for a follow-on task, ask it to read the project context from fenbs first, so the findings from the Notebook phase inform the work.
- Save progress at the end of each session so the next one starts from where this one finished.