The standard advice is to never delete anything. Storage costs nothing, search is good, and you cannot know today what you will want in 2031. Every part of that argument is true, and it still produces archives that people stop opening.
The reason is that a notes system has two jobs and deletion only affects one of them. As storage, more is strictly better — an extra thousand notes cost you nothing you can feel. As a place you look, more is not better, because every browse, every search result page and every folder listing spends your attention on things you did not want. Keeping everything optimises the first job at the expense of the second.
So the question is not "should I delete?" It is "which of these two jobs is this note doing?"
The three-bucket rule
Delete outright — notes whose only property is that they exist.
- Duplicates and near-duplicates. Keep the better one.
- Links saved with no note about why. If you cannot tell from the title what it was for, it is not an antilibrary entry, it is debris.
- Empty templates, test notes,
asdf, notes containing one word you no longer recognise. - Captured tasks that are done or dead. A to-do from March is not a record, it is noise.
- Anything you wrote to avoid deciding. Half of a full inbox is deferred decisions in disguise.
Archive, don't delete — notes that are finished but were real.
- Completed project material. It is context for the next one, and you will want the numbers.
- Meeting notes older than a quarter. Rarely browsed, occasionally decisive.
- Old drafts, once the piece has shipped.
Archiving is the compromise the argument is usually missing: it takes the note out of the place you look while keeping it in storage. Most apps make this a folder or a tag, and most people never use it. It resolves most of the tension on this page.
Never delete — the things you cannot reconstruct.
- Anything about a decision: what was chosen, when, and why. The why is the part nobody writes down and everybody later needs.
- Journal entries. You are a poor judge of which of your own days will matter.
- Notes written in your own words about something you understood at the time. That is the expensive part of progressive summarization, and re-deriving it costs more than the storage ever will.
- Anything with a legal, financial or medical trace.
- Other people's words. Quotes, what someone actually said, exact wording of a commitment.
The two tests
When a note is genuinely ambiguous, two questions settle it faster than any framework.
Could I get this back? A link to a public article is recoverable in ten seconds. Your own reasoning about that article is not recoverable at all. This asymmetry is most of the rule: delete what the internet still has, keep what only you have.
Would I write it again? If the note went missing and you noticed, would you bother recreating it? If not, it was never load-bearing.
Why deletion feels bad
Worth naming, because the discomfort is what actually stops people, not the logic.
There is a real loss-aversion pull: deleting is a definite small loss now against a vague possible gain later, and that trade always feels wrong in the moment. The research literature on digital hoarding describes the same pattern people recognise in physical clutter — accumulation driven by anxiety about future need, with the cost paid in navigability.
There is also a sunk-cost pull. You spent forty minutes on those notes in 2024. Keeping them does not recover the forty minutes.
The counterweight is to notice what the pile costs: every search that returns twelve results where two would do, every browse that ends in a scroll, and the slow erosion of trust in your own system. An archive you do not trust is one you stop consulting, and a notes system nobody consults has already failed in the only way that counts. That is the practical core of PKM mistakes.
Make it a pass, not a mood
Deleting notes one at a time when you happen to trip over them is how nothing gets deleted. Two mechanisms work.
The migration pass. The Bullet Journal's monthly migration is the sharpest version of this idea: you rewrite what still matters onto a fresh page and everything you cannot be bothered to recopy dies quietly. The friction is the filter. Digitally, the equivalent is a filtered view of everything untouched for a year and a fifteen-minute pass over it — a natural add-on to the weekly review, or a quarterly session of its own.
Delete at capture. The cheapest note to delete is the one you never save. A two-second "is this worth keeping?" at the moment of capture removes more junk than any cleanup, without any of the loss aversion — you cannot mourn what you never filed. See the capture habit for where that check belongs, and how to declutter your notes for a full cleanup run.
The default, if you want one
Keep anything you wrote, archive anything finished, delete anything you merely collected. It is not a perfect rule, but it errs in the right direction: it protects thinking and prunes hoarding, which is the trade almost everyone has backwards.
More on the shape of a system that survives this: personal knowledge management and how to organize your notes.