General 12 min read

Your Notes Don't Need a Graph: Backlinks Are a Trap

MMNMNOTE
pkmzettelkastenbacklinksgraph-viewnote-takinglinkingsecond-brain

A backlink records that two notes touch. It does not record why. The graph view is a map of those touches — a topological diagram, not a thought. Nobody has ever thought a new thought by staring at the glowing constellation. The connection that matters is the sentence you write by hand to explain why two notes belong together.

This is not an argument against linking. It is an argument about where the value lives. The system everyone cites as proof — Niklas Luhmann's Zettelkasten — worked on paper, decades before software drew an animated graph.

Luhmann built up "a Zettelkasten of some 90,000 index cards for his research, and credited it for enabling his extraordinarily prolific writing (including about 50 books and 550 articles)." 1 He "linked the cards together by assigning each a unique index number." 2 The link was something he placed, with a reason.

What the graph view promises

The graph view promises to make your thinking visible. Render every note as a dot and every link as an edge, and a structure you could not see before is supposed to emerge — a picture of how your knowledge connects. It is the headline screenshot of modern note tools, the image people share to prove a system is alive.

The promise is seductive because it is partly true. The act of linking does pull weight. Andy Matuschak makes the strongest version of the case: "If we push ourselves to add lots of links between our notes, that makes us think expansively about what other concepts might be related to what we're thinking about. It creates pressure to think carefully about how ideas relate to each other." 3

That pressure is real. A note that forces you to ask what does this connect to is doing cognitive work that an isolated note never asks for. The question is whether the graph captures that work, or merely its residue.

The honest case for the graph

The counter-case deserves a fair hearing before the argument turns. The graph view is not useless, and its defenders are not wrong. Eleanor Konik found a concrete use — orphan detection. "If there are a bunch of orphan things hanging out around the edges of my graph," she writes, "I know that I have not updated my indexes." 4

That is a real job, well done. At small scale the graph stays legible enough to spot a note that drifted loose. Automatic backlinks have a defender too. On the Zettelkasten forum, Ralf Westphal made the fan-in case: backlinks "are useful because they get generated automatically," and they let him "immediately see the fan-in of that zettel (3 other zettels are referring to it)." 5 A note many notes point to is probably one that matters.

So this is not a claim that linking is worthless or that the graph never helps. It is a narrower claim. Hold the concessions in mind, then watch what happens at scale.

Why the graph stops working

The trouble starts where the promise was loudest: at scale. The picture meant to reveal structure dissolves into "a tangled web that's more fun to look at than navigate" past around 200 notes. 6 What looked like a map of your mind becomes a hairball, too dense to read and too pretty to delete.

The deeper problem is not density — it is category. The graph shows that two notes are connected. It cannot show what the connection means. An edge between "spaced repetition" and "deliberate practice" tells you they touch. It says nothing about whether one is an instance of the other, a contradiction of it, or a half-remembered association you made at 1 a.m.

The graph is "a topological map of your connections, not an operational view of your knowledge work." 6 A map of touches is not a map of meaning — and no amount of zooming converts one into the other.

Auto-backlinks fail for the same reason, only worse. A backlink panel lists every note that mentions this one, but it lists them by title, with no reason attached. Sascha, writing at Zettelkasten.de, put the core objection plainly: "Backlinking is just linking notes without connecting knowledge." 7 The list grows. The knowledge does not.

The flaw is mechanical, not aesthetic. Automatic backlinks, Sascha argues, "are bad linking on crack and steroids. Not only do they not provide any link context other than the title, but they are placed automatically." 8

A link you place by hand carries a reason — the half-sentence in your head about why these two ideas belong together. A link a machine places carries nothing but proximity. "With automatic backlinks, this dilution is not even a product of your own choice but the choice of a soulless machine (or your habit, which also has no soul)." 9 The panel fills with connections nobody chose.

The alternative is older than any app. The connection that matters is not the edge — it is the sentence. Sascha's constructive rule is one line: "I always say good linking means that you place any link manually." 10 A manual link is a decision. You ask why two notes relate, then write the reason down.

This is why staring at a graph rarely produces an idea. Following an edge for its own sake is a rabbit hole, not a method. "Following a link for no good reason is not a productive habit," Sascha writes. "In fact, this is the habit that makes the internet a productivity killer." 11

The graph invites exactly that habit. You wander the constellation, admire the shape, and mistake motion for thought. A sentence does not let you wander. It makes you say what you mean.

What to do instead

The practice is smaller than the software suggests. You do not need a panel, a graph, or a plugin — you need a reason and a sentence. Most of the linking apparatus is solving for a picture nobody reads. Three habits replace the whole thing:

None of this requires a tool that draws edges. It requires a tool that stays out of the way while you write the sentence. That is the quiet case for restraint — a minimal note app that keeps your notes as plain Markdown on your own device, where the link is a sentence you wrote and there is no graph to admire. The connection lives in the prose, not in a panel.

Frequently Asked Questions

Is the graph view in Obsidian actually useful?

For a narrow job, yes. At small scale the graph stays legible, and it is genuinely good at spotting orphans you forgot to index. 4 Past roughly 200 notes it becomes "a tangled web that's more fun to look at than navigate." 6 It is a topological map of connections, not a view of your thinking.

Are backlinks actually useful, or just clutter?

Manual backlinks carry a reason; automatic ones carry a title. A link you place by hand encodes why two notes relate. An auto-generated panel, in Sascha's words, is "linking notes without connecting knowledge." 7 The value is in the reason, which only you can write.

Do I need a knowledge graph to think clearly?

No. The Zettelkasten that produced about 50 books and 550 articles ran on index cards linked by hand, with no graph at all. 1 The connection that helps you think is the sentence you write explaining it, not an edge a machine draws between two dots.

Why does my Obsidian graph become a hairball?

Because a graph scales with the number of edges, not the number of ideas. Past around 200 notes the picture is "more fun to look at than navigate." 6 Density is not depth. A graph too dense to read shows you the structure was never in the picture; it was in the notes.

Are automatic backlinks better than manual links?

No. Automatic backlinks are "placed automatically" and add "no link context other than the title." 8 The list grows by proximity, not by meaning. A manual link wins precisely because you had to decide, and writing that decision down is the work that matters.

Should I bother linking my notes at all?

Yes, but link to think, not to decorate. The act of linking "creates pressure to think carefully about how ideas relate to each other." 3 Reserve it for the moment you see a real connection, and write that connection as a sentence.


A backlink records that two notes touch. The sentence you write records why — and that sentence, not the graph, is the thought. If you want to keep your notes somewhere that treats the link as prose you own rather than chrome to maintain, mnmnote.com lives in your browser, in plain Markdown, on your own device.

Footnotes

  1. "Zettelkasten." Wikipedia. https://en.wikipedia.org/wiki/Zettelkasten. Accessed 2026-06-08. 2

  2. "Zettelkasten." Wikipedia. https://en.wikipedia.org/wiki/Zettelkasten. Accessed 2026-06-08.

  3. Andy Matuschak. "Evergreen notes should be densely linked." https://notes.andymatuschak.org/Evergreen_notes_should_be_densely_linked. Accessed 2026-06-08. 2

  4. Eleanor Konik. "It's Not Just a Pretty Gimmick: In Defense of Obsidian's Graph View." September 10, 2021. https://www.eleanorkonik.com/p/its-not-just-a-pretty-gimmick-in-defense-of-obsidians-graph-view. Accessed 2026-06-08. 2

  5. Ralf Westphal (ralfw). "Backlinks in Action." Zettelkasten Forum, November 2020. https://forum.zettelkasten.de/discussion/1487/backlinks-in-action. Accessed 2026-06-08.

  6. Code Culture. "Is Obsidian's Graph View Useful?" 2025. https://codeculture.store/blogs/developer-culture/obsidian-graph-view-useful. Accessed 2026-06-08. 2 3 4

  7. Sascha. "Backlinking Is Not Very Useful — Often Even Harmful." Zettelkasten.de, November 12, 2020. https://zettelkasten.de/posts/backlinks-are-bad-links/. Accessed 2026-06-08. 2

  8. Sascha. "Backlinking Is Not Very Useful — Often Even Harmful." Zettelkasten.de, November 12, 2020. https://zettelkasten.de/posts/backlinks-are-bad-links/. Accessed 2026-06-08. 2

  9. Sascha. "Backlinking Is Not Very Useful — Often Even Harmful." Zettelkasten.de, November 12, 2020. https://zettelkasten.de/posts/backlinks-are-bad-links/. Accessed 2026-06-08.

  10. Sascha. "Backlinking Is Not Very Useful — Often Even Harmful." Zettelkasten.de, November 12, 2020. https://zettelkasten.de/posts/backlinks-are-bad-links/. Accessed 2026-06-08.

  11. Sascha. "Backlinking Is Not Very Useful — Often Even Harmful." Zettelkasten.de, November 12, 2020. https://zettelkasten.de/posts/backlinks-are-bad-links/. Accessed 2026-06-08.