Feed Viewer logo Feed ViewerRead the web on your terms
Curation

A Tagging System for Saved Articles That Survives a Year

I exported my read-later library last spring to check something unrelated and found 1,847 saved articles carrying 240 distinct tags. Of those tags, 118 had been used exactly once.

A laptop showing a read later app with a short sidebar list of tags beside a spreadsheet counting how often each tag was used

I exported my read-later library last spring to check something unrelated and found 1,847 saved articles carrying 240 distinct tags. Of those tags, 118 had been used exactly once. Another 44 were near duplicates of each other: ai and artificial-intelligence and llm, reading and to-read and readlater, plus a tag called misc that had 311 items in it, which is the same as having no tag at all.

That is embarrassing for someone who catalogs things for a living, and it is also completely normal. Tagging feels productive at the moment you save something, because you are summarizing. The problem arrives nine months later, when the tag you need is one you never thought to make and the 240 you did make are in your way. The fix is not more discipline. It is tagging for a different purpose than the one most people tag for.

Why the tag list grows and then stops working

Tags are created at the moment of saving, and at that moment your reason for saving is vivid. You tag it with what the article is about, because that is the obvious thing to write. A week later the reason has faded, and the tag describing the subject is now competing with the full text of the article, which already contains every word on that subject.

Three mechanics do the damage. Synonyms, because nobody remembers which form they chose last time, and most apps will happily create a new tag from a typo. Granularity drift, where a tag starts broad and gains siblings until you have six variants of the same shelf. And the long tail, which in my audit meant 12 tags covered 71 percent of my items and the remaining 228 covered the rest. A list where half the entries are used once is not a vocabulary. It is a pile of notes to yourself.

The symptom to watch for is hesitation. If you pause for more than two seconds deciding which tag to use, your vocabulary is already too large, because you are no longer recognizing a category, you are inventing one.

Tag for what you will be doing, not for what it is about

Here is the reversal that fixed my own library. In a physical catalog, subject headings exist because you cannot search inside a closed book. Every record needs a human-assigned subject or the item is invisible. Your read-later app is not a closed book. It indexes the whole text, including words you would never have thought to assign.

So stop spending tags on subjects. Spend them on the thing search cannot know: your intention. A good tag answers the question of what you will be doing when you want this item again. Those answers are few, repeat constantly, and never appear in the text of the article.

In practice that gives you four kinds of tag worth keeping, and a strict cap on how many of each.

LayerExamplesRuleReview
Actionreply, cite, teach, buy, tryFive or fewer, verbs only, cleared when doneWeekly
Projectp-cataloguing-course, p-house-movePrefixed, finite, retired with a dateWhen the project ends
Formatlongread, video, pdf, paywalledOnly where it changes when you can read itNever, these are stable
Watched namea-linda-holmes, org-internet-archivePeople and institutions you follow on purposeYearly

My working list sits at 14 tags. It has been between 11 and 16 for two years now, and the number is the point. I can see all of them without scrolling, which means I recognize rather than invent.

The project layer is the one people skip and then miss most. A project tag is allowed to be temporary and specific, because it has an end date. When I finished a course last March I renamed its tag with a 2026-03 suffix and stopped adding to it. Those 60 items are now a closed shelf rather than live clutter, and I have never once wanted them merged into something general.

The advice to drop tags and just search is half right

Every few months someone argues that tagging is wasted effort because search exists, and in the version of that argument I used to agree with, folders plus search covers everything. I ran a tagless library for about five months to test it. Search won on three quarters of my lookups and failed badly on the rest, in a way that was predictable rather than random.

Search needs you to remember a word from the item. That works for things you have read. It falls apart for the 400 items you saved and never opened, where you have no memory of the language at all, only a vague sense of why you grabbed it. It fails on anything where the text is not the content, which for me is mostly PDFs of scanned reports and video links. And it cannot group by intention, because nothing in the article says I meant to cite this.

So the honest position is not tags versus search. It is a thin vocabulary covering the three gaps, with search doing everything else. That split is why 14 tags is enough and 240 was worse than none. If your backlog is large enough that none of this feels possible yet, the first pass in clearing a read later backlog is the better place to start, because retagging 1,800 items you no longer want is a waste of an afternoon.

Converting a bloated tag list without a long weekend

I did my own conversion in three sittings of roughly 40 minutes, and most of it was deletion rather than retagging. Export your tag list to a spreadsheet first if your app allows it, because seeing the counts in one column changes what you are willing to throw away.

Delete every tag used once. Not merge, delete. In my case that removed 118 tags and the items underneath them stayed perfectly findable by search, which told me something about how much the tags had been doing. Then merge the obvious synonym clusters down to one surviving form, and pick the shorter one so you type it correctly next time.

What remains will be your top twelve or so, and some of them will be subjects rather than actions. Keep the two or three that genuinely match a standing interest and convert the rest. A tag named productivity became an action tag named try, and that single change made it useful, because now it empties.

One habit to add alongside the cleanup. When you save something, write one short line of why in the notes field rather than reaching for a new tag. Search indexes those notes in most apps, so your own reason becomes findable without inflating the vocabulary. It is also the thing that makes saved highlights resurface usefully later, since a highlight with a reason attached is worth rereading and a bare one usually is not.

Keeping it at this size

Put 15 minutes in the calendar once a quarter. Open the tag list, sort by count, and look at anything that has gained fewer than three items since the last check. Those are candidates for deletion, and deleting them is almost always correct. Retire any project tag whose project ended. That is the whole routine, and mine has never run long.

The test I use after a cleanup is to try three lookups from memory: something I read last year, something I saved and never opened, and something I meant to act on. If all three take under 30 seconds, the system is fine. If the middle one fails, I am missing a project tag rather than a subject one. The rest of our writing on this sits in the curation section.

FAQ

How many tags should a saved article have?

One, usually, and two at most. Items carrying four or five tags in my old library were the hardest to find, because every tag diluted the meaning of the others. If an item needs three tags to describe it, it probably belongs to a project.

Are folders better than tags?

They solve a different problem. Folders enforce one home per item, which is useful for status, so inbox and archive as folders plus a thin tag layer works well. Trying to express subject through nested folders fails the same way a large tag list does.

What about automatic tagging and AI suggestions?

I tested two apps that suggest tags and turned both off within a week. They suggest subjects, which is the layer you do not need, and they create new tags rather than reusing yours, so the list grows while you are not looking.

Should I keep a tag for things I have finished reading?

No, use the app's archive or read state for that. A done tag duplicates a feature you already have and then drifts out of sync with it, which is the worst of both.

WC
Winifred Castellane

Winifred is a digital librarian who tests read-later apps, highlighters and note tools. She writes about reading lists people actually finish and the tools that help.

More posts by Winifred

More in Curation

01doorganico 084goad 3hot8 4465456 after5internprogram ahandfulofblessings ahmadpc ahomcybersecurity allodialresources annodex antiquesandperiodcollectibles aotc asianworldkashmir beedlectrics bloggertemplatefree bluefeatheragency bodyweightsolution campaign08blog carfixadvisor catstreetcollective cbstvsports cnfanspurchasing committeeof500yearsofdignityandresistance covertopsdjs cs-kaiyuangame cuestacomunicaciontotal cxoutlet dancebetgame deskgearhub destinyvillastay detectmagazine di33 diamondsteelbuildings donmo dyanadropship emarsysfinance emimattl escapechan eskitarih evautoplay everythingweddings-orlando farmzone foxstudiosscratch gamevui123 gleasongrade greetingsday hacklabasu haskell iw-kaiyun jagoanproperti jilidskripsi joltbusinessfinance juegodepokergratis juegospucca kaeru-festival latestmehndidesigns lestanzediottaviano litemagazines localcommunityhero lustrsazan mariashebets menswheels newwebstechs1 nikkit ootheday oxigenvpn plurfitness prefetial probudgetcarpetcleaning psychicmrsmartin pyaten qvantor rejoignezlogis rexzbot riselayouts ruchin rwxqsgd5 sakaltime shibiaowang shop-bot skrtips spotlesshaven strcoding tetapatrium topselller uneplanteartificielle vipertechcr webtech1223411 wo-kygaming yamaha-mekar yeknava yoybuysheets zenunwa yamaha-mekar jilidskripsi ankertoto ankertoto ankertoto