A PhD breaks note systems in a way coursework never does, for one reason: the time horizon. You are taking notes in year one that you will need in year five, on a question that will have changed three times in between, and nobody will remind you where you put anything. Undergraduate note-taking is optimised for an exam in eight weeks. Doctoral note-taking is an archive problem.
That is why the honest answer here is not an app. It is three layers, and most people who finish end up with something like them regardless of which brands they chose.
The three layers
| Layer | Job | Reasonable choices |
|---|---|---|
| References | Every paper, cited correctly, forever | Zotero, or Mendeley / EndNote if your department insists |
| Thinking | Your ideas, linked, in your own words | Obsidian, Logseq, or plain markdown |
| Writing | Producing the actual chapters | Word, LaTeX/Overleaf, or Scrivener |
Trying to collapse these into one tool is the most common mistake, and it usually fails at the reference layer. Citation management is a solved problem with a specialist tool, and a general notes app will not do it as well.
Layer one: Zotero, and it is not close
Get this right in month one. Zotero is free, open source, has a browser connector that captures a paper and its metadata in one click, generates bibliographies in any style, and — critically — will still exist and still export in year five.
The rules that save you later:
- Capture every paper you even skim, with the PDF attached. Storage is cheap; re-finding a half-remembered paper is not.
- Fix the metadata at capture time. Ten seconds now, versus a bibliography audit the week before submission.
- Use one library with tags, not deep folder trees. Your topic structure will change; tags survive that, folders do not. See tags vs folders.
- Turn on syncing and check the backup. People lose PhD libraries. See how to back up your notes.
Zotero is where papers live. It is not where your thinking lives, and using it as a note archive is the second most common mistake.
Layer two: where the thesis actually gets built
This is the layer that matters and the one people neglect for two years.
The core discipline is that reading notes and your own ideas are different objects. A note that quotes Foucault and a note that argues something are not the same kind of thing, and merging them is why year-three literature reviews are so painful to write. Keep literature notes — what the source says, in your words, with a page number — separate from permanent notes, which are your claims, standing alone.
Obsidian is the common choice, for good reasons: markdown files you own, links and backlinks, and a mature Zotero integration that pulls citations and annotations into your vault. A PhD is exactly the timescale where plain text notes pay off, because you will outlast at least one software fashion.
Where it fails: setup is a genuine time sink, and the plugin rabbit hole has consumed real months of real doctorates. Configure it in an afternoon and stop.
Logseq suits people who think in daily notes and outlines rather than documents; Notion is better at project tracking than at thinking, and its search degrades at the scale a PhD reaches; Roam popularised the model and costs more than the alternatives now.
The method matters more than the tool. A linked, atomic archive built over four years is the closest thing to a shortcut that exists at writing-up time — see Zettelkasten and smart notes. The system's whole promise is that when you sit down to write chapter four, the argument already exists in fragments and you are assembling rather than starting.
Layer three: writing, and the LaTeX question
If your field uses LaTeX, use it, and use Overleaf if collaboration matters more than local control. If your supervisor comments in Word, you will be in Word regardless of your preferences, and fighting that is not a battle worth the time.
Scrivener deserves a mention for humanities theses specifically. Its strength is holding a 300-page document as movable pieces rather than one file, which suits the way long arguments actually get restructured. Its weakness is that export to a formatted thesis is fiddly, and it is a writing tool rather than a thinking one.
The transition from layer two to layer three is where most of the pain lives. Whatever you choose, test the export path early — notes to draft to formatted chapter — rather than discovering the problem in the final month.
What about Anki
Worth it for language requirements, technical vocabulary, methods and formulae. Largely not worth it for the conceptual content of your own field, which you will learn through use. Spaced repetition is powerful and it is aimed at retention of discrete facts, which is a smaller part of doctoral work than it is of medical school. See the best notes app for medical students for the case where it dominates.
The failure modes, honestly
Collecting instead of thinking. A 2,000-item Zotero library and 40 notes is the classic year-two position, and it feels like progress. It is the collector's fallacy at doctoral scale.
Rebuilding the system every six months. Each migration costs a fortnight and produces nothing. Pick adequate tools, then stop shopping.
Notes with no argument in them. Summaries of papers are cheap and nearly useless in year five. What you will want is what you thought, and where you disagreed. Write those down at the time — see how to read a research paper.
No record of dead ends. The experiment that failed, the framework you abandoned and why: these are invisible in your notes and will be asked about in your viva. Keep a dated research log. See how to keep a work journal.
Where ClairMind fits
ClairMind is a fast, private, phone-first notes app with AI over your own notes. In a PhD stack its honest role is the capture layer — the objection that occurs to you in a seminar, the connection between two papers that arrives on the walk home, the methods worry at 11pm. Those thoughts are the ones most likely to be lost and often the most valuable, and neither Zotero nor Obsidian is reachable fast enough to catch them.
It does not manage citations and it is not a thesis-writing tool, so it sits alongside Zotero and your vault rather than replacing either. See also the best notes app for research and how to take research notes.
The minimum viable stack
Zotero for every paper from day one. A folder of markdown notes where each note is one idea in your own words. Your department's writing tool. A dated log of what you tried and what failed.
That is a complete system, it costs nothing, and it will still be readable when you are defending it.