Recall captures your snapshots, but searching them returns nothing new — or the indexing indicator sits at the same percentage for days. The timeline grows while search stays frozen in the past. That is a stuck index: the background job that turns raw captures into searchable content has stalled, and everything downstream of it is starved.

Indexing is what makes Recall Recall. Without it, you have screenshots with no memory attached — search cannot match queries to what was on screen. A stuck index is one of the most functionally damaging Recall faults.

This guide gives you six fixes in technician order: confirm the stall, restart the indexing pipeline, clear its blockages, and rebuild it cleanly when repair is not enough. Work through them in sequence.

Why this happens

  • Indexing service crashed silently: The background service that processes snapshots can die without notification. Captures continue (a different component), but nothing new gets indexed.
  • Corrupted index database: The semantic index can corrupt — especially after forced shutdowns — and a corrupt index stalls the writer rather than crashing outright.
  • Storage exhaustion mid-index: The index needs working space. A full Recall quota or nearly-full system drive freezes the indexing writer in place.
  • A single toxic snapshot: Occasionally one malformed capture — a truncated image, a broken metadata blob — chokes the indexer, which retries it forever instead of skipping ahead.
  • Conflicting search index state: Recall’s semantic index and Windows’ own search index can interfere when one is being rebuilt while the other holds locks on shared components.
  • Paused by power policy: Aggressive battery-saver or sleep policies can suspend the background indexing task indefinitely, so the percentage never advances while the machine is “on” but idle.

How to fix it

Work through these in order — diagnose first, then act.

Fix 1: Confirm the stall is real, not just slow

Indexing a large backlog legitimately takes hours, and interrupting it mid-run can cause the very corruption you are trying to fix. Verify the stall before you touch anything.

  1. Note the current indexing status or percentage, then use the PC normally for at least 4 powered-on hours without sleeping.
  2. Check again — any movement at all means the indexer is working through a backlog, not stuck.
  3. Confirm the machine is not in battery-saver mode and that sleep is not interrupting background work; plug in the charger during this observation.
  4. If the figure has not budged after a full day of powered-on time, proceed — the stall is genuine.
  5. Meanwhile, search behavior that finds nothing even for recent items is covered from the search side in the search not finding anything guide .

Fix 2: Restart the Recall indexing service

A crashed or wedged indexing service is the most common cause, and a clean restart clears the transient state that wedged it — without touching your snapshots or your existing partial index.

  1. Press Win + R, type services.msc, and find the Windows Recall-related services.
  2. Stop the indexing-related service, wait a full 30 seconds, then start it again.
  3. Open Task Manager’s Details tab and confirm the Recall indexing process is actually running and consuming some CPU — zero activity after a restart means it died again instantly.
  4. Give it one hour of powered-on time and check whether the percentage moved.
  5. If it stalls at the exact same point again, a single toxic snapshot may be the blocker — continue to Fix 4.

Fix 3: Free the storage the indexer needs

An indexer with nowhere to write cannot advance. This fix clears both the quota ceiling and system-drive pressure that freeze the indexing writer mid-run.

  1. In Settings > Privacy & security > Recall & snapshots, check storage usage against your allocation — if it is at the ceiling, delete the oldest snapshot range.
  2. Open Settings > System > Storage and ensure the system drive has several gigabytes free; clear temporary files if not.
  3. Restart the indexing service again (Fix 2) after freeing space, since a writer frozen by exhaustion does not always resume on its own.
  4. Monitor for two hours — a space-starved indexer usually surges forward quickly once it can write again.
  5. Microsoft’s Windows support hub documents Recall’s storage behavior if you want to size things properly going forward.

Fix 4: Isolate and clear a toxic snapshot

When indexing always stalls at the same point, one malformed capture is likely choking the pipeline. Removing the snapshots around the stall point lets the indexer skip past the damage.

  1. Estimate which time period corresponds to the stall point — if indexing froze at 60%, the problem snapshot is roughly 60% through your timeline’s date range.
  2. In Settings > Privacy & security > Recall & snapshots, use Delete snapshots to remove a narrow time range around that estimated point.
  3. Restart the indexing service and watch whether it advances past the previous stuck percentage.
  4. If it does, the deleted range contained the toxic capture — a small, acceptable loss of history for a working index.
  5. Do not delete your entire history for this; a surgical range deletion is enough, and a full reset remains a last resort.

Fix 5: Rebuild the Windows search index alongside Recall’s

Recall’s semantic search shares infrastructure with Windows Search, and a damaged system-wide search index can hold locks or feed bad state into Recall’s pipeline. Rebuilding both clears the interference.

  1. Open Control Panel > Indexing Options (search for it in Settings) and open Advanced.
  2. Click Rebuild to regenerate the Windows search index from scratch — this takes a while and runs in the background.
  3. While that runs, restart the Recall indexing service once more so both rebuilds proceed together.
  4. Leave the PC powered on and plugged in; avoid sleep until both indexes report healthy.
  5. This is also a good moment to rule out unrelated faults — for instance, live captions failing on YouTube is a captioning issue rather than an indexing one, and x64 emulation problems on ARM live in the compatibility layer, not the search stack.

Closing

A stuck Recall index is almost always a wedged service, exhausted storage, or one bad snapshot — all repairable without losing history. Restart the pipeline, give it room to write, surgically remove the toxic capture if there is one, and rebuild only when repair fails. With the index moving again, the timeline becomes memory instead of just pictures.

Frequently asked questions

What does it mean when Recall indexing is stuck?
Recall’s semantic index — the database that makes your timeline searchable — has stalled partway through processing snapshots. Captures continue, but search cannot find anything new until indexing resumes.
How long should Recall indexing normally take?
Initial indexing after setup or a reset can take several hours on a large backlog, running quietly in the background. If the percentage has not moved for 24 hours of powered-on time, it is stuck, not slow.
Will restarting indexing delete my snapshots?
No. These fixes restart or rebuild the index, a derived database built from your snapshots — your snapshot store and timeline stay intact. Only a full store reset would touch your history.
Can low disk space stall Recall indexing?
Yes — the index needs working space to write processed data. A nearly-full system drive or a Recall store at its quota ceiling are two of the most common reasons indexing freezes mid-run.