# YouTube AI Slop Shorts Dataset

A dataset of **4,224 YouTube channels** and **139,136 Shorts video IDs** labeled for AI-generated content detection.

## Overview

This dataset was created to help researchers, developers, and content moderators identify AI-generated ("slop") content on YouTube Shorts. It contains channel-level and video-level labels for AI slop detection.

## Stats

| Metric | Value |
|---|---|
| Channels | 4,224 |
| Shorts Videos | 139,136 |
| Channels with labels | 948 |
| Videos with labels | 390 |

## Format

```json
[
  {
    "channel": "@channelName",
    "labels": {
      "slopChannel": ["SLOP", "SUSPECTED_AI"]
    },
    "shorts": [
      {"videoId": "abc123"},
      {"videoId": "def456", "aiSlop": 1}
    ]
  }
]
```

## Fields

### Channel level
| Field | Description | Count |
|---|---|---|
| `channel` | YouTube channel handle | 4,224 |
| `labels.slopChannel` | AI slop label (`SLOP`, `SUSPECTED_AI`) | 889 |
| `labels.badChannel` | Flagged as bad | 26 |
| `labels.badChannelCustom` | Custom bad flag | 49 |
| `labels.safeChannel` | Verified safe | 1 |
| `labels.goodChannel` | Verified good | 3 |
| `labels.goodChannelCustom` | Custom good flag | 1 |

### Video level
| Field | Description |
|---|---|
| `videoId` | YouTube video ID |
| `aiSlop` | AI slop flag (242 videos) |
| `slop` | Slop flag (141 videos) |
| `safe` | Safe flag (8 videos) |

## Usage

### Python
```python
import json

with open("youtube_ai_slop_dataset.json") as f:
    data = json.load(f)

# Get all AI slop channels
slop_channels = [ch for ch in data if "slopChannel" in ch.get("labels", {})]
print(f"AI Slop channels: {len(slop_channels)}")

# Get all labeled videos
for ch in data:
    for video in ch["shorts"]:
        if "aiSlop" in video or "slop" in video:
            print(f"https://youtube.com/shorts/{video['videoId']}")
```

## Use Cases

- Training AI content detectors for YouTube Shorts
- Browser extensions for flagging AI-generated content
- Research on AI slop prevalence on social media
- Content moderation tools

## License

This dataset is provided for research and educational purposes.

## Contributing

If you'd like to contribute labels or report issues, please open an issue or pull request.