PBS uses Amazon Bedrock’s GenAI to tag 700,000 assets in months, enhancing how viewers search and discover content.
PBS has integrated Amazon Bedrock’s GenAI into its digital platforms to overhaul how users discover shows and educational content. Built in collaboration with the AWS Generative AI Innovation Center, the new AI-powered search engine enables intuitive, content-aware exploration across the PBS App and PBS LearningMedia.
The challenge was vast: PBS needed to organize decades of content spanning 700,000 media assets. Manual metadata tagging would have taken years. Instead, using Bedrock’s models—particularly Anthropic’s Claude Sonnet—PBS automated the entire process in under six months. The system learned to interpret transcripts, descriptions, and context, creating thousands of tags that allow viewers to search by themes, tones, or emotions, such as “heartwarming” or “historical drama.”
This generative AI-driven cataloging improved search precision and discovery speed, helping users uncover lesser-known programs while maintaining PBS’s editorial integrity. Early testing revealed that AI tagging matched or outperformed manual curation while cutting processing time from hours to minutes. According to PBS’s Director of Digital Innovation, Mikey Centrella, this shift “turned metadata creation from a bottleneck into a growth engine.”
Beyond faster search, the project showcases how public institutions can safely adopt GenAI at scale. Running within PBS’s secure AWS cloud, the system ensures privacy while enabling future AI expansion—such as personalized recommendations for children’s content. The partnership demonstrates how GenAI enhances legacy archives into dynamic, user-centric experiences, helping PBS meet evolving viewer expectations in a crowded streaming landscape.
