> ## Documentation Index
> Fetch the complete documentation index at: https://docs.bookmarkjar.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Cleanup & Duplicates

> Learn how to find and remove duplicate bookmarks using AI-powered cleanup

## Overview

The Cleanup feature helps you maintain a tidy bookmark collection by identifying and removing duplicate or similar bookmarks. It uses multiple detection strategies—from simple URL matching to AI-powered semantic analysis—to find duplicates you might have missed.

<Frame>
  <img src="https://mintcdn.com/bookmarkjar/PGRJQMP62AYJ_W7M/images/cleanup-create.png?fit=max&auto=format&n=PGRJQMP62AYJ_W7M&q=85&s=7c749d4a2ff68ab8f415b66665f5ae98" alt="Cleanup Review Interface" width="2078" height="1700" data-path="images/cleanup-create.png" />
</Frame>

<Note>
  Nothing is deleted until you explicitly approve. All cleanups go through a review process where you can see exactly what will be removed.
</Note>

## How It Works

1. **Create a Review** — Configure detection methods and filters
2. **Analysis** — The system scans your bookmarks and groups duplicates
3. **Review** — Browse grouped duplicates and choose which to keep
4. **Approve** — Confirm the cleanup to soft-delete duplicates
5. **Restore** (24h) — Accidentally deleted? Restore within 24 hours

## Creating a Cleanup Review

Navigate to **Dashboard → Cleanup** and click **New Review**.

### Detection Methods

<AccordionGroup>
  <Accordion title="Exact Duplicates (Free)">
    Finds bookmarks with identical or nearly identical URLs. This catches:

    * Same URL with different formatting (`http://` vs `https://`)
    * URLs with/without `www.` prefix
    * URLs with trailing slashes removed
    * URLs with fragments (`#section`) removed
  </Accordion>

  <Accordion title="Domain Duplicates (Free)">
    Finds bookmarks from the same domain with similar content. Useful for catching:

    * Same page bookmarked with different tracking parameters (`?utm_source=...`)
    * Mobile vs desktop URLs (`m.twitter.com` vs `twitter.com`)
    * Domain aliases (`x.com` vs `twitter.com`)
  </Accordion>

  <Accordion title="AI Semantic Analysis (Premium)">
    Uses AI embeddings to find bookmarks about the same topic, even if URLs are completely different. Perfect for:

    * Same article syndicated across multiple sites
    * Similar content with different titles
    * Bookmarks covering the same topic from different sources
  </Accordion>
</AccordionGroup>

### Filters

Fine-tune which bookmarks are analyzed:

| Filter                   | Description                                                  |
| ------------------------ | ------------------------------------------------------------ |
| **Exclude Pinned**       | Skip bookmarks you've pinned (default: on)                   |
| **Exclude Protected**    | Skip bookmarks marked as protected (default: on)             |
| **Older Than X Days**    | Only include bookmarks older than a specified number of days |
| **Not Viewed in X Days** | Only include bookmarks you haven't viewed recently           |
| **Domains**              | Include or exclude specific domains                          |

## Reviewing Duplicates

After analysis completes, you'll see your duplicates organized into **groups**. Each group contains bookmarks that the system considers duplicates of each other.

<Frame>
  <img src="https://mintcdn.com/bookmarkjar/PGRJQMP62AYJ_W7M/images/cleanup-review.png?fit=max&auto=format&n=PGRJQMP62AYJ_W7M&q=85&s=d3a948b0cd47fbd98c81af7cde1861a0" alt="Cleanup Review Interface" width="2152" height="1946" data-path="images/cleanup-review.png" />
</Frame>

### Understanding Groups

Each group shows:

* **Keeper** (marked with a crown) — The bookmark that will be preserved
* **Duplicates** — Other bookmarks in the group that will be deleted

### Changing the Keeper

Don't agree with the suggested keeper? Click any bookmark in the group to set it as the new keeper. The system intelligently picks keepers based on:

* Cleanest URL (no tracking parameters)
* Most complete metadata
* Most recent save date

### Skipping Bookmarks

If you don't want to delete a specific duplicate, you can mark it as "Keep" to exclude it from deletion while keeping the rest of the group's actions intact.

## Approving a Cleanup

Once you've reviewed all groups:

1. Click **Approve Cleanup** at the bottom of the review
2. Confirm the action in the dialog
3. Duplicates are soft-deleted (moved to trash)

<Warning>
  Soft-deleted bookmarks are permanently removed after 24 hours. Use the **Deleted** section to restore any mistakes.
</Warning>

## Using Chat for Cleanup

You can also trigger cleanup through the AI chat. Just describe what you want to clean up:

* "Find duplicate bookmarks"
* "Clean up my Twitter bookmarks"
* "Remove bookmarks I haven't viewed in 30 days"
* "Find similar articles about React"

The chat will create a cleanup review based on your request.

## Tips for Best Results

<CardGroup cols={2}>
  <Card title="Start with Exact Duplicates" icon="bullseye">
    Run exact duplicate detection first—it's fast, accurate, and catches the most obvious duplicates.
  </Card>

  <Card title="Use Domain Duplicates for Social Media" icon="share">
    Social media URLs often have tracking params. Domain duplicate detection cleans these up nicely.
  </Card>

  <Card title="AI Analysis for Deep Cleaning" icon="sparkles">
    Premium users: Enable semantic analysis to find topically similar bookmarks across different sources.
  </Card>

  <Card title="Exclude Important Bookmarks" icon="shield">
    Pin or protect bookmarks you never want to delete before running cleanup.
  </Card>
</CardGroup>

## Frequently Asked Questions

<AccordionGroup>
  <Accordion title="Can I undo a cleanup?">
    Yes! Deleted bookmarks go to a soft-delete state for 24 hours. Visit **Dashboard → Cleanup → Deleted** to restore them.
  </Accordion>

  <Accordion title="What happens to old reviews?">
    Cleanup reviews expire after 7 days if not acted upon. Creating a new review also marks previous pending reviews as expired.
  </Accordion>

  <Accordion title="Is semantic analysis available on the free plan?">
    No, AI semantic analysis requires a Premium subscription. Free users can still use URL-based and domain-based detection.
  </Accordion>

  <Accordion title="How accurate is the duplicate detection?">
    * **Exact duplicates**: 100% accurate (same URL hash)
    * **Domain duplicates**: \~95% accurate (URL normalization + content similarity)
    * **AI semantic**: \~85-95% accurate (depends on content similarity threshold)
  </Accordion>
</AccordionGroup>