Backlinks Were Supposed to Resurface Your Notes. They Do Not, and Here Is What Does

Backlinks arrived in note-taking apps a few years ago with a promise attached: link your notes to each other, and your older thinking will rise back up at exactly the moment it becomes relevant. Roam made them the centre of the product, Obsidian and Logseq adopted them, and the graph view — the spiderweb of dots and lines that renders your notes as a network — became the screenshot everyone posted to prove their system was alive. I spent a real stretch of weeks building links on purpose, and the feature did almost none of what it was sold as doing. The notes that resurfaced were the ones I already knew existed.

Here is the mechanical problem, and it is hiding in the name. A backlink is a link you create now, pointing from one note to another. To create it, you have to be holding both notes in your head at the same moment — the new thought and the old one it connects to. That is precisely the moment you do not need the link, because you are already thinking about the old note. The link earns its keep later, when you have forgotten the old note and something should bring it back. But a link you wrote in the past can only point at notes that were on your mind in the past. It cannot point at the note you forgot, because you forgot it before you could link it.

The result is a graph that is a map of your awareness, not your memory. The notes with the most backlinks are the ones you think about constantly — your current project, the idea you are excited about this month, the framework you quote in every conversation. The note with zero links, sitting in a corner unconnected, is exactly the note a retrieval system is supposed to rescue, and it is the one the graph is structurally blind to. Backlinks are a popularity contest, and in note-taking the note you need is rarely the popular one.

The graph view makes this worse by dressing it up as insight. It looks like knowledge — hundreds of nodes, a dense constellation, the satisfying sight of a system that has become a thing — and almost none of that visual density is usable. You cannot ask a graph a question. You cannot search it, because it is a picture, not an index. The moment you have a real retrieval need — “where did I write down that thing about onboarding?” — you leave the graph and type into the search box, which is the confession that the beautiful network was never the tool you were going to use. The graph is a dashboard for a system that works somewhere else.

So what actually resurfaced my old notes? Full-text search, and it was not close. When you search across every word you have ever written, the note you forgot comes back the instant you type even a fragment of what it was about. The catch is that search only works as well as the words inside the note, which shifts the real problem one step earlier: retrieval is decided at the moment of writing, not at the moment of searching. A note titled “meeting” or “thoughts” or “misc” is invisible to search no matter how good the search engine is, because the words that would find it were never written down. This is the same point I keep landing on — a note has a shape, and a findable note has a title and a first sentence that carry the retrieval keys.

The title is the single highest-leverage line in the whole system, and backlinks have nothing to say about it. A good title is not a summary, it is the phrase you will think of in six months when you need the note again. When I write a note I now ask one question: “what would I type to find this later?” and I make that exact phrase the title. It is a small discipline with an outsized return, because it front-loads the retrieval work into the thirty seconds when the thought is still fresh, instead of asking a future version of me to reverse-engineer what past me meant.

The second thing that worked was deliberately writing more prose into each note, not less. The Zettelkasten tradition teaches short, atomic notes with dense linking, and I ran that faithfully for months before I noticed the pattern: my shortest, most densely-linked notes were the ones I never reopened, because there was too little text for search to bite onto. A single paragraph with one link is a note that only exists if you already know it exists. Three paragraphs that spell out the idea in plain words are findable from a dozen different search terms. Links connect notes to each other; words connect notes to future you, and it is the second connection that actually pays rent.

This is where backlinks still earn a small place, and I want to be precise about it rather than throw them out. A backlink is a genuinely useful navigational cue in two narrow cases: when you are actively working inside a cluster of related notes and want to move between them, and when a link lets you walk backward from a note to discover what else referenced it. Neither of those is retrieval from cold memory. Used as a browsing aid inside a note you are already in, backlinks are pleasant. Used as a system for resurfacing things you have forgotten, they are the wrong tool, and the confusion between those two jobs is why the feature got so much credit it never earned.

The honest summary is that I over-invested in the glamorous half of the system and neglected the boring half, and the boring half was the whole game. Linking is a way to organise what you can see; search, plus a title worth typing, is the way to recover what you cannot. If I were starting over I would spend almost none of my energy on links and all of it on writing notes that a stranger could find by accident — because the person who finds your note six months from now is a stranger to the person who wrote it. If the links themselves are the part of your system you keep fiddling with, it is worth asking whether you built a filing cabinet or a way to actually get notes back out.