How to search your X bookmarks
Premium gets you a search box. Here is what it can and cannot find.
X Premium puts a search box over your bookmark list that matches words in the saved posts. On a free account there is no bookmark search at all: the main search bar covers public posts, not your private saves.
The box works well on the case it was built for, which is a phrase you actually remember. It fails on four cases it never covers, and one of those four accounts for most of the searches people describe as broken.
Using it
Open Bookmarks and use the search field at the top of the list. It matches against the text of the posts you have saved, so the useful queries are distinctive phrases rather than topics. Search for a phrase you remember reading, not for the subject you would have filed it under.
A trick that works better than it should: search for the author's handle. You often remember who posted something when you have entirely forgotten what it said.
Why a search misses something you know you saved
The most common miss is a retrieval ceiling, not a broken search box. That ceiling, the observable 800 to 1,000 horizon, is the subject of Is there a limit on X bookmarks?. Search only covers what the list can still reach. Three reasons, in the order they are likely:
- It is below the retrieval horizon. Older saves stop being served, and search cannot match what the list will not load. This is by far the most common cause on an old account.
- The post is gone. Deleted posts, and posts from suspended or newly private accounts, cannot be matched on text nobody can serve any more.
- The words are in an image. A screenshot of text is a picture. Nothing in the bookmark search reads it.
That last one catches people constantly, because a large fraction of what gets bookmarked on X is screenshots.
No operators, no folders, no dates, nothing inside links
It is a keyword match over post text, and that is genuinely all. In particular:
- No search operators. The from:, since:, and filter: operators that work in X's main search are not part of the bookmark search box. Typing them searches for the literal characters.
- No filtering by folder. Search covers the list; folders are a separate view. You cannot ask for "posts in Reference that mention pricing".
- No date range. You cannot narrow to last March.
- Nothing inside links. If you saved a post whose only content is a link to an article, the article's text is not searchable. Only the post's own words are.
You are searching the sentence, not the thing the sentence pointed at.
On a free account: three mediocre workarounds that work
There is no built-in search, and the workarounds are all mediocre. In fairness, they do work:
- Find-on-page. Open bookmarks on a desktop browser, scroll until the section you want has loaded, then use the browser's find. It only searches what is loaded, so scroll first.
- Export and search the file. Get a CSV or JSON out with an extension and search that. More effort up front, much better afterwards, and you end up with a backup.
- Search X itself. If you remember roughly what the post said and it is public, finding it again through normal search is sometimes faster than finding it in your own bookmarks.
That last one is a fair summary of how weak private-save retrieval is: it is often quicker to re-find a public post than to locate your own copy of it.
What repeated failed searches are actually telling you
Repeated failed searches are a signal, and the signal is usually not "I need a better query". Three patterns, with what each one is really telling you:
- You find recent saves and never old ones. That is the retrieval ceiling and no amount of searching gets under it. Export, and stop treating the bookmark list as storage.
- You cannot remember enough to search at all. You are saving things you have no real relationship with. That is not a search problem, it is a save-rate problem, and searching harder will not fix a collection you never engaged with in the first place.
- You find it, and it turns out you did not need it. Worth noticing how often this happens. If most successful searches end in a shrug, the collection is costing you more in hunting than it returns.
A spreadsheet beats the product at this
If you export to CSV or JSON you get a strictly better search than the product offers, and it is worth knowing what that unlocks:
- Everything at once. No retrieval ceiling. The file is the file.
- Date ranges. Sort by the timestamp column and look at a period.
- By author. Filter the handle column and see every save from one person, which is often how you actually remember things.
- Regular expressions, if you are comfortable with them, which turns "something about a number and a percentage" into a query.
A spreadsheet is a worse reading experience and a much better retrieval tool. For a collection you are mining rather than reading, that is the right trade.
Worked examples
Concretely, here is what works and what does not when you are hunting for something:
| You remember | Try |
|---|---|
| A distinctive phrase from the post | Search that exact phrase. This is the case the box is built for and it works well. |
| Who posted it | Search the handle including the @. Often faster than remembering the content. |
| Roughly the topic | Search two or three words you would expect in the post itself, not the category you would have filed it under. The box matches the post's words, not your mental label. |
| It was a screenshot | Search the caption, since the image text is not indexed. If there was no caption, scroll. |
| It was ages ago | Search will probably fail. This is the retrieval ceiling, not your memory. |
The fix is upstream of search
If you are regularly failing to find saves, the fix is upstream of search:
- Quote-post instead of bookmarking for things you will genuinely want again. Your own posts are searchable by anyone, you included.
- Copy the link somewhere with real search at save time, if it matters.
- Keep the collection small enough to see. A list under the horizon is fully searchable. Every save above it degrades retrieval for everything below.
The last point is the one that generalises. Search quality on a personal collection is mostly a function of collection size, and the only lever that never stops working is subtraction.