I've been chasing the dream of a unified workspace for months. One tool that handles research, synthesis, and recall without forcing me to alt-tab between Notion, Readwise, and a janky Markdown editor. NotebookLM felt like the answer as Google's AI research assistant that lets you upload sources, ask questions, and generate insights from your own material. The interface is clean, the audio overviews are genuinely impressive, and the promise is simple: dump your research in, get intelligent analysis out.

But NotebookLM still misses fundamentals that stop it from being a true daily driver. Every time I think I can consolidate my workflow, I hit the same walls. There are some foundational gaps that undermine the entire experience. Here's what breaks.

Why I keep trying to make NotebookLM my everything app

The promise is irresistible

NotebookLM does something no other tool quite manages: it treats your sources as the ground truth. Instead of hallucinating answers from the general web, it analyzes what you've uploaded (be it PDFs, Google Docs, website links, or YouTube transcripts) and builds responses solely from that material. This constraint makes it weirdly endearing in ways ChatGPT and Perplexity aren't.

The Audio Overview feature converts your sources into a podcast-style conversation, which sounds gimmicky until you're commuting and absorbing a research paper through synthesized dialogue. The inline citations link back to specific passages, so you can verify claims without scrolling endlessly.

And the interface stays minimal with no feature bloat, no gamification, just a clean workspace where you interact with your material. It feels like the beginning of something transformative. Except it keeps stopping short.

It can't export anything useful

Your insights stay trapped in Google's ecosystem

I spent two hours synthesizing findings from six product design articles in NotebookLM. The chat history was pristine, flowing with organized arguments, cited sources, and clear takeaways. Then I needed to share it with my team in Slack.

There's no export. You can copy-paste individual responses, but citations don't transfer as links. The formatting breaks. If you've generated multiple threads exploring different angles, there's no way to package them into a portable document. The Audio Overview? You can download the audio file, but not the transcript or the sources it references. This isn't just a missing feature but a key structural oversight.

Research tools live or die on portability. Obsidian lets you export Markdown. Notion gives you PDFs and HTML. Zotero integrates with Word and LaTeX. NotebookLM keeps everything locked inside its own environment, which makes it feel less like a research assistant and more like a walled garden with great conversation but no exit. I've started screenshotting chat threads just to preserve formatting. That's a workflow regression.

Cross-notebook memory doesn't exist

Each notebook starts from zero

NotebookLM organizes projects into separate notebooks, which makes conceptual sense. You can create one notebook for your thesis, another for client work, a third for personal learning. But notebooks can't talk to each other. If you've uploaded overlapping sources or asked similar questions across projects, NotebookLM won't surface those connections. I uploaded a paper on cognitive load theory into my "Learning Design" notebook and a UX research report into my "Product Strategy" notebook. Both cited the same foundational study. NotebookLM treated them as isolated facts in separate universes.

There's no cross-referencing, no "you explored this concept here" prompt, no unified knowledge graph. This fragmentation kills long-term value. Tools like Obsidian and Roam thrive on backlinks and emergent connections between notes. Readwise Reader builds a personal search engine from everything you've highlighted. NotebookLM forces you to remember which notebook contains which insight, then manually recreate context every time you start fresh. It's amnesiac by design.

The irony is that Google has already solved this problem elsewhere. Gmail's search spans your entire archive. Google Drive surfaces related documents automatically. But NotebookLM treats each project like a hermetically sealed chamber, which feels like a step backward from Google's own ecosystem strengths.

Task capture feels like an afterthought

You can't build a system around follow-ups

Mid-conversation, NotebookLM will surface a genuinely useful insight like a gap in your research, a contradictory claim between sources, a thread worth exploring. The natural next step is to capture that as a task: "Investigate Smith's counterargument" or "Find more recent data on this trend." NotebookLM has no task management layer. You can't flag responses for follow-up, create a running list of open questions, or tag conversations by status. Some tools integrate with task managers. For example, Notion has databases, Capacities has objects, and even Apple Notes comes equipped with checklists. NotebookLM just scrolls endlessly forward.

I tried using the chat itself as a task tracker, typing "TODO" before questions I wanted to revisit. It became unmanageable within days. The interface isn't built for retrieval. Instead, it's optimized for generation. So I'm back to maintaining a parallel system in Todoist, manually copying tasks out of NotebookLM, which defeats the entire point of a unified workspace.

The gap becomes especially painful during iterative research. When I'm exploring a topic over weeks, I need to track what I've learned, what contradicts it, and what still needs verification. NotebookLM gives me the synthesis but no scaffolding to build on it systematically.

Citations lack the depth research actually needs

Source attribution is surface-level

NotebookLM's inline citations are better than most AI tools. They link directly to passages in your uploaded sources which is legitimately helpful. But they don't go deep enough for serious research. There's no page number or paragraph reference you can cite in a formal paper. If you upload a 40-page PDF, NotebookLM might attribute a claim to "Source 3" with a vague highlight, but you still need to manually hunt down the exact location for your bibliography. Zotero auto-generates citations in APA, MLA, Chicago, whatever you need. Notion's AI can reference block-level links. NotebookLM gives you a general pointer and calls it done.

It also doesn't track provenance across conversations. If you ask the same question in two different chat threads, NotebookLM might pull from different sources without flagging the inconsistency. There's no citation history, no "you relied on this source three times but ignored this contradictory one" transparency. For exploratory thinking, that's fine. For rigorous research, it's a liability.

I tested this by asking NotebookLM to compare two competing theories across different sessions. It gave me coherent answers each time, but drew from different subsets of my uploaded sources without acknowledging the variance. A proper research tool would flag that inconsistency and let me decide which framing to prioritize.

Long-term recall depends on you remembering notebooks exist

Discovery breaks down after three projects

I have 11 notebooks in NotebookLM right now. I can't remember what half of them contain. There's no global search across notebooks, no recent activity feed, no smart suggestions like "this new source relates to your work from two months ago." Each notebook is a sealed container that only surfaces when you manually open it.

Compare this to Readwise, which resurfaces highlights from everything you've read based on spaced repetition. Or Notion, where linked databases and backlinks create serendipitous rediscovery. NotebookLM assumes you'll maintain perfect mental overhead of every project you've ever started, which is absurd once you pass the five-notebook threshold.

I've started abandoning notebooks instead of revisiting them because the friction is too high. The insights are still there, technically. But if the tool doesn't help me rediscover them, they might as well not exist.

The most frustrating part is that NotebookLM has all the raw material to make this work. It knows what sources I've uploaded, what questions I've asked, and what patterns emerge across conversations. But it refuses to connect those dots unless I explicitly open the right notebook and remember to look. That's not the best knowledge management workflow. Rather, NotebookLM becomes storage with a chat interface.

These aren't edge cases

NotebookLM needs to decide what it wants to be

Here's the fundamental tension: NotebookLM markets itself as a research assistant, but it's missing the infrastructure that research actually requires. Export portability, cross-project memory, task tracking, rigorous citations, and intelligent recall aren't power-user features. They're table stakes for anyone doing sustained intellectual work. Without them, NotebookLM stays stuck in the demo phase, impressive in 20-minute sessions but frustrating over weeks of real use.

Google has the resources to fix this. The underlying AI is strong, the interface design is solid, and the source-grounded approach genuinely differentiates it from ChatGPT and Perplexity. But until NotebookLM addresses these foundational gaps, it's a tool I keep trying to love and keep walking away from. I want it to replace my fragmented workflow. Right now, it just adds another fragment. Fix the scaffolding, and NotebookLM could genuinely become the everything app it promises to be.