21,238 YouTube Videos with Multimodal Metadata and Clickbait Labels

YTClickbait21K is a human-annotated dataset of 21,238 YouTube videos for clickbait detection research. It includes video metadata like titles, descriptions, and thumbnails, along with binary clickbait labels from three annotators per video. The dataset was uploaded by Md. Minhazul Islam to figshare on April 9, 2026.

Use Cases

• Train binary classifiers for clickbait detection based on video titles and descriptions.
• Develop multimodal models that combine text features (title, description) with visual features (thumbnail).
• Study annotator agreement and confidence in human labeling tasks using the provided annotation metadata.
• Analyze relationships between clickbait labels and video performance metrics like view and like counts.

Strengths

• Contains 21,238 unique video records, providing a substantial sample size.
• Each video has three independent human annotations and a derived consensus label.
• Includes 26 columns of multimodal features, such as text, numerical metadata, and image URLs.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Data may reflect temporal and source bias inherent to the specific YouTube videos collected.

Provenance

Source

figshare

Collection Method

Collection of YouTube video metadata combined with multi-annotator labeling.

Time Range

Upload dates for videos are included, but the overall temporal coverage is not specified.

Freshness

Last updated 2026-04-09 08:35:40; freshness should be verified.

Geography

Spatial coverage is not specified.