Finding and searching

Search basics

Type in the dashboard search box (press / to focus it). Results update as you type.

• The text is split on spaces. Every word must match (AND).
• A plain word matches if it appears in the bookmark's title, address or folder path . Matching ignores case and is a substring match, so git matches github.com .
• Word order does not matter.
• Search works on the current view only: the dashboard's whole library, narrowed by the active filter chip and any selected tags.

Operators

Add an operator to a word to look in one place only. There is no space after the colon.

Operator	Matches	Example
title:	The title contains the text	title:invoice
url:	The full address contains the text	url:/blog/
path:	The folder path contains the text	path:reading
host:	The host name contains the text	host:docs.
domain:	The registrable domain contains the text	domain:example
health:	The link-health status equals the value. Values: healthy, redirected, broken, server-error, unreachable, unknown, plus the shortcuts bad (broken, server-error or unreachable) and good or ok (healthy or redirected)	health:bad
is:	A cleanup flag: duplicate, untitled, old, collision (or title-collision)	is:duplicate
age:	age:old or age:stale: older than 730 days	age:old

Combine them freely, for example domain:example is:duplicate health:bad or path:reading title:guide.

What search does not do

• There are no quoted phrases. "exact phrase" is treated as separate words, quote marks included.
• There is no tag: operator. Use the tag chips (below).
• An operator with no text after it ( title: ) is ignored, and unknown is: values have no effect.
• health: only matches bookmarks that already have a recorded link-health result. Until you run a scan, nothing is "broken" or "healthy" (see Link health ).
• The dashboard search is exact. Fuzzy matching is a popup feature (see below).

Filter chips

The chips under the search box apply a cleanup filter on top of your search. Only one is active at a time, and the choice is remembered between visits.

Chip	Keeps bookmarks that…
All	No cleanup filter
Duplicates	Share a page address with at least one other bookmark
Untitled	Have no real title (empty, untitled, new bookmark or (Untitled bookmark))
Old	Were added 730 or more days ago
Collisions	Use a title that is also used for a different address

The number on each chip counts the whole library. Press Esc (with the cursor outside the search box) to clear the search text first, then the active chip. See the Escape ladder.

Ordering results

Search narrows the list; it does not rank it. In the dashboard, the Sort visible bookmarks control beside the grouping selector orders what is left by title, address, newest, oldest or folder path (see Dashboard tour).

Tag chips

When at least one bookmark has a tag, a Tags: strip appears above the list. It shows your 12 most-used tags plus any tag you have switched on. Click a chip to require that tag; click again to remove it. With several tags on, a bookmark must have all of them. Clear removes the tag filter. Clicking a tag on a row does the same as clicking its chip. Tags are described in Organizing.

Fuzzy search (popup only)

If the popup finds no exact match and your text has no operators, it ranks approximate matches: an exact substring scores highest, then letters that appear in order. Title matches are weighted most, then address, then folder. A line reading No exact matches — showing fuzzy results tells you it happened. The dashboard never switches to fuzzy matching, so an empty dashboard result means no exact match.

How Bookmark Scope decides two URLs are the same

Duplicate detection, the popup's page scope, the toolbar badge, link-health result keys and the duplicate preview all use one definition of a "page":

1. Only http and https addresses are understood. Other schemes ( file: , chrome: , javascript: ) are never matched or checked.
2. The host name is lower-cased. The path is kept case-sensitive .
3. A trailing slash is dropped ( /docs/ equals /docs ).
4. http and https are different pages.
5. A non-default port is part of the identity ( :8443 ); a default port such as :443 on https is not.
6. The query string ( ?a=1 ) is kept by default. Turn on Ignore query string in page matching to drop it.
7. The fragment ( #section ) is ignored by default. Turn off Ignore hash fragment in page matching to keep it.

Both options live in Settings. Changing them re-evaluates every duplicate count and the popup scope.

Examples

Each row compares two addresses using the default options (query kept, fragment ignored). Same page? is what the Duplicates filter and Page scope use.

Address A	Address B	Same page?	Same host?	Same domain?
https://example.com/docs	https://example.com/docs/	Yes	Yes	Yes
https://example.com/docs	https://EXAMPLE.com/docs	Yes	Yes	Yes
https://example.com/docs	https://example.com:443/docs	Yes	Yes	Yes
https://example.com/docs	http://example.com/docs	No	Yes	Yes
https://example.com/docs	https://example.com:8443/docs	No	Yes	Yes
https://example.com/docs	https://example.com/Docs	No	Yes	Yes
https://example.com/docs	https://www.example.com/docs	No	No	Yes
https://example.com/docs?a=1	https://example.com/docs?a=2	No (Yes if query is ignored)	Yes	Yes
https://example.com/docs#top	https://example.com/docs#end	Yes (No if fragments are kept)	Yes	Yes
https://docs.example.co.uk/a	https://shop.example.co.uk/b	No	No	Yes
https://alice.github.io/a	https://bob.github.io/b	No	No	No

The last two rows show the Public Suffix List at work: co.uk is a public suffix, so example.co.uk is one site; github.io is also listed as a suffix, so every *.github.io project is its own domain.