An AI notebook is not a document archive
An AI notebook is a workspace you fill with sources for one piece of thinking, and it is very good at that. A household archive is a place records live permanently, arrive continuously, and get found by several people years later. Using either for the other job is where the frustration comes from.
What is an AI notebook actually shaped for?
You bring sources to a question. A lease and the two amendments; a syllabus and eleven papers; a manual, a warranty and the email chain about the repair. The tool reads what you gave it, answers from those sources, and cites which one it took each answer from. That last part is the reason people trust it for thinking work — the answer points back at the page.
Everything about the shape follows from “one question at a time”. Google's help pages list up to 50 sources per notebook and 100 notebooks, with 50 chat queries a day (checked July 2026). Those are the proportions of a project, and they are the right proportions for one.
Where does that shape stop fitting household records?
Four places, and none of them are faults — they are simply what a project-shaped tool does when handed a decade of paperwork.
Records arrive forever, projects end
A notebook is complete when you have gathered what you need. A household archive is never complete: something lands every week, and it has to land somewhere without anyone deciding which project it belongs to.You have to know which notebook to open
Splitting years of paperwork across notebooks recreates the folder problem with a nicer interface. The question “when did we last renew the tenant insurance” only works if you already know which notebook holds it.Some official documents will not go in
the one people hit firstGoogle lists copy protection as a reason an import fails (checked July 2026). Banks and government departments issue protected PDFs routinely, so a portion of exactly the documents a household most wants to keep will refuse to load.Households are plural
Records are needed by whoever is dealing with the thing that day. An archive has to answer to more than the person who built it, and without that person being available to explain the structure.
How do the two compare, side by side?
| AI notebook | Household archive | |
|---|---|---|
| Unit of work | A notebook per question | One place, always |
| Lifespan | As long as the task | Years — decades for some documents |
| Capacity | Up to 50 sources per notebook, 100 notebooks | Whatever the household accumulates |
| How things get filed | You choose the sources for each notebook | By what the document is, without a decision per file |
| Who uses it | Usually the person doing the thinking | Anyone in the household who needs a document |
| Best question to ask it | “What do these sources say about X?” | “Where is the thing, and what did it say?” |
Which one do you need?
Three questions usually settle it. Is the set of documents finished, or does it keep growing? Will anyone besides you need to find something in it? Will you be looking for it in a year, when you no longer remember what you were working on when it arrived?
Growing, shared, and looked at years later means you want an archive. Bounded, personal, and read this month means a notebook is the better tool, and reaching for an archive would be overkill. Plenty of people want both — a permanent home for the records, and a notebook opened when something needs thinking through.
What does the archive side look like?
Sparky Fetch is built for the second job. Photos and PDFs go in by web upload, by forwarding them to the Telegram bot, or through a synced Google Drive folder; each one gets OCR and lands in a category without anyone naming it; and everyone in the household can search the same archive or simply ask it a question. It is in private, invite-only beta, so the honest offer is a look rather than a sign-up — the live demo is a fictional family's paperwork, no account needed.
A fuller feature-by-feature comparison lives on the NotebookLM alternative page. If you are trying to decide between several tools generally, choosing a document tool lists the questions worth asking any of them.
Frequently asked questions
How many documents can a Gemini Notebook hold?
Google's help pages list 100 notebooks, up to 50 sources in each, and up to 500,000 words per source, with daily limits of 50 chat queries and 3 audio generations (checked July 2026). A single upload can be up to 200 MB with no page limit. Fifty sources is generous for a research project and small for a decade of household paperwork.
Is an AI notebook a safe place for personal documents?
Google's help page states that your data will never be shared by Gemini Notebook, and links to a page explaining how it protects that data (checked July 2026). Whether that is the right home for your records is a different question from whether it is safe: the shape of the product is a workspace you fill for a task, not a permanent place with a filing structure your household navigates for years.
Why can't I add some of my PDFs?
Google lists three reasons an import fails: the source is over the 500,000-word limit, the file is over 200 MB, or the original PDF is copy-protected (checked July 2026). The last one catches people out with bank statements and official documents, which are often issued with protection on.
So when is an AI notebook the right tool?
When you have a bounded set of material and a question you want to think through against it: a contract you are trying to understand, a stack of research papers, a policy document and its amendments. You gather sources for the task, work through them, and the notebook has served its purpose when the task is done.
And when do you need an archive instead?
When the material never stops arriving, more than one person needs to find things in it, and you will be looking for something specific in three years without remembering which project it belonged to. That is a different shape: everything in one place, organised by what each document is, and reachable by whoever needs it.
Different jobs. Both worth doing. ♡
Gemini Notebook limits and behaviour are from Google's Gemini Notebook help centre FAQ, checked July 2026 — the product Google launched as NotebookLM, whose help pages now carry the new name. Google changes limits often, so read the live page before relying on a number here. Details about Sparky Fetch describe current behaviour during private beta. Nothing here is affiliated with or endorsed by Google.