Deep Dive Nº 03 · Study Tools

Turn your own lectures into a tutor

Two Google tools now read only the notes you feed them: NotebookLM and Gemini study notebooks. Point them at your slides and readings and you get citations, a podcast, and a quiz that adapts to what you keep getting wrong.

Most revision advice quietly assumes you have already done the hard part: condensed a term of lectures into something you can actually test yourself on. That is the bit almost nobody finishes. A generic chatbot does not help much here, because it answers from the whole internet, confidently and often wrongly, with no idea what your lecturer actually said. The interesting shift in 2026 is a pair of tools that do the opposite. They read only the material you give them, and they hand the work of turning it into revision back to you in a form you can practise from.

The two worth knowing are NotebookLM and Gemini study notebooks, both from Google and both built on its Gemini models. They overlap, but they solve different halves of the same problem: one turns your sources into study assets, the other turns them into a tutor that quizzes you. Used well, together they close the gap between a folder of PDFs and a revision plan.

The point is not to read your notes back to you. It is to make you retrieve them, which is the one thing that actually moves information into memory.

The grounded notebookAn assistant that only reads your sources

NotebookLM starts empty. You upload sources: lecture PDFs, slide decks, a reading list, your own messy notes, even a YouTube lecture or a pasted transcript. Everything it then says is grounded in those documents and nothing else. Technically it uses retrieval-augmented generation, which means it fetches the relevant passages from your uploads before answering rather than reaching for whatever it half-remembers from training.

The practical payoff is citations you can check. Every answer carries small numbered chips, and clicking one opens the source document scrolled to the exact passage, with the sentence it drew on highlighted. That single feature changes how you can trust it: instead of wondering whether a claim is real, you verify it in seconds against the paragraph in your own lecture notes. Ask it "what did week 7 say about elasticity", and it answers from week 7, footnoted.

From reading to listeningPodcasts, mind maps and flashcards from your PDFs

The feature that made NotebookLM briefly famous is the Audio Overview: it generates a surprisingly natural podcast-style discussion between two AI voices talking through your material, available in 80+ languages and in formats from a quick brief to a longer deep dive. For a dense topic you half-understand, hearing it argued out loud on the walk to campus is a genuinely different way in.

The rest of the Studio panel is arguably more useful for exams. From the same uploads it will generate:

The mental model: NotebookLM is a machine for converting passive sources into things you can be tested by.

The tutor that quizzes backStudy notebooks and Guided Learning

Gemini's study notebooks, rolling out through 2026, take a further step. You upload your syllabus and notes, and instead of producing assets, the app sets a short diagnostic quiz to find your weak spots. From your results it builds a plan of bite-sized lessons, each with its own practice quiz, and a dashboard that sorts more than 100 learning objectives into strengths, focus areas and not-yet-started. As you answer, the plan recalibrates: get something right and it moves on, get it wrong and it circles back.

Underneath sits Guided Learning, the mode Google launched in the Gemini app in September 2025. Rather than handing over a finished answer, it walks you through a problem step by step, asking questions and checking you have followed before moving on. Google's own framing is blunt about why:

Google's stated aim

"The idea is to help you learn, not just provide the answer. Knowing how to arrive at an answer is more critical than the answer itself." Standardised-test prep (SAT now, with ACT, GRE and others due in summer 2026) and richer diagrams are on the same roadmap.

Why testing beats rereadingThe learning science underneath

None of this is magic. It works, when it works, because it nudges you into two techniques cognitive scientists have backed for decades: active recall (retrieving an answer from memory rather than rereading it) and spaced repetition (revisiting material at growing intervals). Rereading feels productive because the words look familiar. Familiarity is not memory, and that gap is exactly where students overestimate how ready they are.

~80%
Recalled a week later after repeated retrieval practice, in Karpicke and Roediger's 2008 study
~36%
Recalled by the group that only restudied the same material, same study
g = 0.50
Mean benefit of retrieval practice over restudy across 159 comparisons (Rowland, 2014)
g = 0.61
Benefit of practice testing over other methods in a later meta-analysis (Adesope et al., 2017)

Those are large, replicated effects. A quiz drawn from your own lecture forces retrieval; a podcast that talks the concept through, then asks you to recall it later, spaces the repetition. The tools are useful mainly to the degree they get you doing these two things, not because the AI is clever.

Turning a lecture into revisionA workflow that actually holds

A concrete loop that fits a single module:

One module, one week
StepToolWhat you get
Dump the sourcesNotebookLMSlides, readings and notes in one grounded notebook
Map itNotebookLMA mind map plus a study guide to see the shape
Test yourselfNotebookLMFlashcards and a quiz from your own material
Fix weak spotsStudy notebookDiagnostic quiz, then adaptive lessons on the gaps
Revise on the moveAudio OverviewA podcast of the week to re-listen to before class

The order matters. Generate assets first, then spend most of your time being tested by them. The temptation is to stop at the shiny podcast, which feels like studying but is still passive.

Read before you relyThe honest caveats

Grounded does not mean infallible, and it is worth being blunt about the limits:

The takeawayThe mental model to keep

Think of these tools as a way to shorten the distance between raw material and self-testing, not as a shortcut around the testing itself. NotebookLM turns your sources into things that can quiz you; Gemini study notebooks turn your weak spots into a plan; and the reason either helps is the oldest finding in the study-skills literature, that retrieving beats rereading. The AI is the assistant. The recall still has to be yours.

✦ End of issue ✦

Get the next one in your inbox

One AI tool, one prompt, three quick wins, every week. Free, and under 3 minutes to read.

Subscribe Free →