Synthiz vs NotebookLM: which should you choose?

What each tool is for

NotebookLM, built by Google, works in notebooks. You drop documents into one — PDFs, text, links, transcripts — then question that specific corpus. The model answers only from what you supplied, which sharply reduces the risk of invention, and it cites the passages it leans on. It also derives formats from the notebook: summaries, study guides, timelines, and conversational audio renditions. It is a deep-analysis tool whose scope is explicitly drawn by you, every time.

Synthiz works across time rather than within a session. You hand it a link — an article, a YouTube video, a podcast, a PDF — or you subscribe to an RSS feed, a channel, a show; it syncs on its own, hourly, every six hours or once a day. Anything audio or video is transcribed, summarised, and its concepts extracted before you open the app. Every statement carries timestamped citations and reopens the transcript at the exact second. Cross-source analysis puts sources side by side to show agreements, tensions and blind spots, and Thoughts hold your own reflections, anchored to those sources.

NotebookLM answers 'what is in these twenty documents'. Synthiz answers 'what have I accumulated over a year, and where did that idea come from'.

Comparison table

Criterion	NotebookLM	Synthiz
Working scope	A closed notebook you upload	A continuous base fed by subscriptions
Deep analysis of a given corpus	Excellent, core of the product	Good, but designed for a stream
Citations back to sources	Yes, to documents in the notebook	Yes, timestamped into the transcript
Search	Semantic, but limited to the open notebook	Hybrid, semantic and lexical, across the whole base
Browser extension	Capture into a notebook depending on the version	One-click capture and search of your own base from the page
Automatic monitoring (RSS, channels, podcasts)	No	Yes, hourly, 6-hour or daily sync
Podcast and video transcription	Depends on the source you upload	Yes, systematic
Audio rendition of the corpus	Yes, conversational format	No
Personal notes anchored to sources	Notes inside the notebook	Thoughts linked to sources
Cost	Free base plan	Free account, paid plan at €9 per month

When to choose NotebookLM

NotebookLM is excellent, and on its own ground it is hard to beat.

The closed corpus, first. You have thirty research PDFs to understand before Friday, or a project's entire documentation to absorb. Drop them in, ask your questions: the answer does not leave the perimeter and it cites its passages. That strict boundary is a guarantee a tool working over an open base cannot offer as cleanly.

Then the price. The base plan costs nothing, which remains a massive argument for occasional use or for a student.

Finally the quality of what it produces. The derived formats — study guide, timeline, and above all the conversational audio version of the corpus — are well made and have no equivalent in Synthiz. It should also be said plainly that NotebookLM has semantic search over its notebook: this is not a crude keyword engine, and on an uploaded corpus it finds an idea phrased differently very well.

When to choose Synthiz

The difference is not analysis quality, it is time horizon. A NotebookLM notebook is a session: you assemble it, you exploit it, you move on. Continuous research does not stop — six newsletters a week, three YouTube channels, two podcasts. Nobody uploads that by hand every Monday. Synthiz's subscriptions absorb the stream and process it before you open the app.

The second gap is long-term retrieval. Questioning twenty documents you just uploaded is easy: you know what is in them. Finding, eight months later, the exact source of an idea you noted in passing, inside a base that grew while you were not watching, is a problem of a different nature. Synthiz fuses a dense semantic score over bge-m3 embeddings with a lexical full-text score, and does not rely on a fixed similarity threshold: relevance is measured by how far a document sits from the distribution of its own query, mean + z*stddev with an absolute floor. On a measured corpus of 948 embeddings, a relevant query — 'feudalism', five genuinely concerned documents — and a query matching nothing at all were only 0.09 apart; a fixed 0.65 cutoff returned nothing to a user who was in fact well supplied. The consequence is twofold: you find things without recalling the words, and when the base holds nothing on a subject, the answer is a flat zero results rather than three off-topic documents.

The third is the point of entry. The Synthiz Chrome extension saves in one click the page, the YouTube video whose transcript it pulls automatically, or the podcast you are listening to; more importantly, it queries your base without leaving the page, showing the excerpt around the term you searched for rather than a list of results. 'Do I already know this?' gets settled mid-video, which a notebook you have to open in another tab cannot do.

Verdict by profile

Student preparing an exam on a defined corpus: NotebookLM. Free, precise on what you upload, and its study guides are immediately useful.

Professional who has to absorb a one-off dossier: NotebookLM. The closed notebook is exactly the right shape for a file with a beginning and an end.

Analyst, journalist, sector watcher: Synthiz. The stream is continuous, it has to be captured without intervention, and the source of an idea must stay findable months later.

Content creator: Synthiz for the accumulated base and the timestamped citations, NotebookLM occasionally when a subject calls for digging into a closed corpus.

Pricing moves: check each vendor's site before deciding.