Link health: checking, scanning and repairing

Link health is the only feature that contacts the sites you bookmarked. It is off until you ask, it asks permission first, and it never edits anything by itself.

The short version

A check sends a lightweight request to each bookmarked address from your own browser and records what came back.
You must grant Chrome's optional site access for http and https pages first. If you say no, nothing is requested.
Results are kept in a local cache for seven days.
A result is a hint, not a verdict. Sites block bots, require sign-in and rate-limit, which can look like "broken".

What the statuses mean

Status shown	When it is assigned	What it usually means
Healthy	The server answered with a success status and no change of address	The page loaded. This does not prove the content is the same as when you saved it.
Redirected	The request was redirected, or ended on a different page than the one you saved (differences in the `#fragment` are ignored)	The page moved. You may be able to repair the bookmark. A different `#fragment` or host letter case alone does not count as a move.
Broken	HTTP status 400 to 499, shown with the code	Not found (404), gone (410), but also sign-in required (401), forbidden (403) and rate-limited (429). Check before deleting.
Server error	HTTP status 500 or higher	The site had a problem. Often temporary.
Unreachable / Timed out	No response: DNS failure, refused connection, no network, or no answer in 8 seconds	The site may be gone, or you may be offline.
Unchecked	No result recorded yet	Run a check.
Check failed (skipped)	The request was skipped by policy, aborted, or produced no response	Chrome Web Store addresses and non-web addresses are skipped.

How the check works

The extension asks for the page's headers with a HEAD request, following redirects and bypassing the HTTP cache.
If the server answers 405 Method Not Allowed , it tries a GET instead, with a 5-second limit.
The normal limit is 8 seconds per address.
The classification looks at the status code and whether the final address changed. It does not read or judge the page content.

The permission model

Bookmark Scope declares http://*/* and https://*/* as optional host permissions. They are not granted when you install.

You start a check. Choose Inspect visible, Inspect selected, Inspect this group, or start a resumable scan.
Chrome asks. If access has not been granted, Chrome shows its own prompt for site access. Read it. Bookmark Scope needs the broad pattern because it contacts whichever sites you bookmarked.
You decide. If you decline, you get Health scans need temporary site access. Permission denied, so the scan was not run. and no request is made.

Revoking access

The prompt wording says "temporary", but Bookmark Scope does not revoke the permission for you afterwards. To remove it, open chrome://extensions, choose Details for Bookmark Scope and change its site access there. Chrome controls those screens and their exact wording.

Enabling the scheduled scan in Settings also asks for the same permission when you save.

Running a check from the dashboard

Choose what to check. The check covers your selected bookmarks if any are selected; otherwise everything visible after filters. Duplicate addresses are checked once.
Press Inspect. The button reads Inspect visible or Inspect selected. While it runs it shows Stop inspect 12/80. Up to four addresses are checked at a time.
Watch the results appear. Rows gain a health badge (for example Healthy 200 or Broken 404). Hover a badge for a redirect target or an error note. The Health panel and the score update as results arrive.
Stop any time. Press the same button again. Results already received are kept: Inspection stopped after 12/80 links… Partial results were kept.

When a run finishes you get a summary such as Health scan finished for Entire library: 70 healthy · 3 redirected · 4 broken · 1 server · 2 unreachable. A scan is recorded in the Activity panel.

To find what a scan flagged, search health:bad (broken, server error or unreachable), health:redirected or health:good.

Resumable scans

For long lists, use Library tools → Resumable scan. It keeps its progress in storage so closing the page does not lose it.

The dialog states its scope (selected or visible) and Unique supported URLs to check .
Start new scan creates the job and begins. If an unfinished job already exists, tick Starting a new scan replaces the previous progress first.
Pause scan , Resume scan and Cancel scan do what they say. A canceled job cannot be resumed; start a new one.
The progress line reads Completed / total / remaining followed by a status: No saved scan , Ready to start , Scanning , Paused , Canceled or Completed .
Addresses are checked one at a time. A job can hold between 1 and 50,000 addresses, and only one job is saved.

It is paused, not running in the background

A resumable scan only runs while its dashboard tab is open. Closing the dialog, or the tab, pauses it. Reopen Library tools; a Saved scan progress button shows how far it got, and Resume scan continues. If a request was in flight when it stopped, that address may be checked again, but it is never counted twice.

Scheduled background scans

If you want link checks without opening the dashboard, enable the scheduled scan in Settings.

It is off by default . Enabling it prompts for site access; if you decline it stays off.
The interval is 1 to 90 days (default 7). It is driven by a Chrome alarm, so it only runs while the browser is running, and timing is approximate.
Each run checks up to 250 bookmarks, one at a time. It remembers where it stopped and continues from there next time, so over several runs it works through the whole library instead of re-checking the same first 250.
If, after a run, some bookmarks have newly turned broken, server-error or unreachable, you get one notification: Bookmark Scope: new broken links found — N bookmarks now appear broken. Open the dashboard to review and clean up.
Settings shows Last scan: …. N bookmarks currently look broken. This number counts every cached result, including ones from manual scans.

A scheduled scan never deletes, edits or moves anything.

Repairing redirects

When a check finds that a bookmark now lands on a different address, you can update the bookmark to the final address.

Use Repair redirected in the selection bar (selected bookmarks), Repair redirect in a row's menu, the same button in the Stats panel recommendations, or Library tools → Repair redirected (all visible bookmarks that qualify).
You are asked Repair redirected bookmarks? with the count. Confirm with Update bookmarks .
Only the address changes. Title, folder, position, ID, date and tags are kept. The bookmark's stale health result is dropped so it can be checked again.

Be careful

Repairing a redirect cannot be undone from Bookmark Scope. There is no Undo toast for it. Open the detail panel first and read Redirects to: a redirect to a sign-in page, a cookie notice or a homepage is not a repair. Export a snapshot if you are unsure.

The health cache

Results are stored locally, keyed by the normalised page address (so the URL-identity options affect them).
A result older than 7 days is ignored. The cache holds at most 5,000 results; if it overflows, the oldest are trimmed.
Library tools → Clear cached health data (also in the list's More actions menu) asks Clear cached health data? and forgets all results. Bookmarks are untouched.
If you clear the cache while a scan is running, results from that earlier scan cannot repopulate it.

Practical tips

Check a group or a domain first ( Inspect this group ) before the whole library.
Re-check anything flagged Broken 403 , 401 or 429 in your browser before deleting it.
A flagged Unreachable result while you were offline or on a restricted network is not evidence the site is gone.
Intranet or localhost bookmarks may only work from certain networks; results reflect where your browser is.