I subscribe to maybe thirty newsletters. I read maybe five of them properly. The other twenty-five just pile up in my Gmail until I open one with a subject line I half-remember and feel vaguely guilty about not reading two months ago. Then I tried routing them into NotebookLM (Google’s free AI notebook that reads whatever you upload and answers questions across it) and the dynamic changed. The archive became something I could actually search, query, and cross-reference, not just a guilty unread count.
The trick is one two-step Zap (a Zapier automation, two app connections wired together) on the free tier. Gmail fires when a labeled email arrives, and Google Drive saves the body as a text file in a folder that NotebookLM watches. From there, the value compounds. After a few weeks, I could ask “what apps has this newsletter recommended more than once” and get a real answer instead of scrolling through archived issues.
Why I did not just leave them in Gmail
Gmail search is good at finding a specific email you remember. It is not good at “what did this newsletter say about backup tools across the last four issues.” NotebookLM is built for that shape of question. It treats a folder of documents as a corpus and answers against the whole thing, so the longer you let the workflow run, the more useful it gets.
I had been doing the manual copy-paste thing for a while. It worked for a week. Then I missed two issues, and the whole archive fell behind. Manual is fine for a newsletter you read occasionally, but a weekly subscription needs something that does not depend on me remembering.
Building the two-step Zap
Zapier’s free plan caps at 100 tasks a month with one trigger and one action per Zap. A weekly newsletter uses about four tasks a month, so the budget is not the bottleneck. The trigger app is Gmail, with the event “New Labeled Email.” I created a Gmail label called “Newsletters” and applied it to the senders I wanted to capture. The Zap fires only when an email hits that label. Nothing else in my inbox gets touched.
The action app is Google Drive, with the event “Create File From Text.” Map the email body into the content field, point the destination at a folder NotebookLM can watch, and turn the Zap on. Every new issue lands in Drive on its own. NotebookLM will auto-ingest new files in a watched folder, so there is nothing to add by hand.
There is a gotcha. Zapier’s “Create File From Text” preserves HTML markup in the body. The NotebookLM source ends up with broken formatting and stray links, which makes the underlying reading rough but does not break the queries. If you want clean text, you would need a third step (a formatter or a small Zapier formatter hook) that strips HTML first. I left it. The queries still work, and I can always open the original email when I need the formatting.
Prompts that actually earn the archive
After a few weeks, I stopped using NotebookLM for summaries and started using it for pattern questions. The prompts that worked best:
- “Across every issue in this notebook, list every tool or recommendation the writer has mentioned more than once. For each one, note how many issues mentioned it and whether the writer’s opinion changed over time.”
- “Compare how the writer’s take on [topic] has shifted across the last four issues. Quote the strongest disagreement.”
- “Pull the most contrarian or surprising claim from this archive, with the issue and date it appeared.”
Simple summary prompts are a waste. The value is in queries that span issues. A newsletter’s real job is the pattern across months, not the contents of any single issue.
A few habits that made the archive more useful over time:
- One notebook per topic, not per sender. Two newsletters on similar themes work better in a single notebook than separate ones. You get cross-source comparison for free.
- Re-add the source after major edits. NotebookLM caches the original version. If the writer publishes a “corrections” addendum to a past issue, you have to drop and re-add it.
- Ask for quotes, not summaries. “Quote the strongest claim” returns something you can paste into a doc. “Summarize” returns mush you have to re-read.
- Pin the best prompts. Save the queries that actually worked into a small text file in the same Drive folder. When you start a new notebook, paste them in and tweak.
Studio output as a passive review
NotebookLM has a Studio panel that can generate an Audio Overview (a two-host AI podcast reading the sources) or a mind map from a notebook. The mind map is the one I use most. It surfaces how the newsletter’s recurring themes actually connect before I have to commit to a deeper prompt. The Audio Overview works if I want to catch up on a month of issues on a long drive. Both work better when the notebook is on a single topic than when it is a grab bag.
Scaling to more newsletters
The same two-step Zap works for any other newsletter. Duplicate it, change the Gmail search query to the new sender, point it at a new Drive folder, and add the folder to NotebookLM. I run four Zaps now, one per newsletter I care about. Each notebook is its own corpus, so cross-referencing is opt-in. I tried merging them into a single mega-notebook once. The signal-to-noise dropped. Smaller notebooks with clearer topics are better.
Trade-offs
The free tier is the limit. Hundred tasks a month sounds like a lot until you stack four weekly newsletters plus a couple of dailies. The cap will start to matter. If you hit it, the cheapest fix is to upgrade or move to Make (the Integromat successor) which has a different free-tier shape. None of the alternatives strip HTML cleanly without an extra step, so the broken-formatting issue is universal.
NotebookLM still does not auto-detect changes inside an existing source file. If the source file is edited in Drive after NotebookLM ingests it, you have to re-add it. In practice this does not matter for a newsletter archive, where every issue is a new file. It would matter for a wiki or shared doc.
The privacy story is also Google’s. Every issue you pipe in is uploaded to a Google service. If the newsletters you care about include anything sensitive, that is a real cost. Most of mine do not, so it has been a non-issue for me. Your math will be different if you handle confidential material through email.
The migration took about forty minutes for the first Zap, including creating the Gmail label and the Drive folder. The easier-than-expected part was the Gmail trigger. The harder-than-expected part was figuring out which “Create File” event on Drive actually mapped the email body. Zapier has three similar events and only one of them does what you want.
If you read more than two newsletters a week and you keep losing track of them, set this up. Start with one newsletter you actually like, let it run for two weeks, and ask NotebookLM one cross-issue question. If the answer surprises you, you will keep the workflow. The whole point is that after a few months you stop having a newsletter graveyard in your inbox and start having a research notebook you can query.