Skip to content
Sean Bettsv2.0.2
Back to Building
research toolArchived

YouTube SDG Analysis

An LLM-based analysis system that scored 2.5 million YouTube videos against an ethical media framework based on the UN Sustainable Development Goals.

Sole developer · January–December 2025
YouTube SDG workflow: collect channel metadata, video information and transcripts; analyse content, mission and themes; then map directly evidenced themes to relevant Sustainable Development Goals. Structured results support channel-level analysis.
In practice

What it does

  1. Collect video transcripts and channel information for analysis.
  2. Use LLMs to analyse video content against an existing ethical scoring framework.
  3. Map relevant themes to the UN Sustainable Development Goals and produce structured scores.
  4. Bring video-level results together to support channel analysis.
Behind the work

The challenge

I worked with a startup that sources ethical media opportunities for brands and wanted to extend its offering to YouTube. I devised and built a way to use LLMs to analyse video transcripts against its existing ethical scoring framework, based on the UN Sustainable Development Goals. The work involved translating that framework into an automated analysis process and scaling it to 2.5 million videos.

What I learned

We shut the project down after YouTube cut off the transcript API we were using. The system had reached scale, but losing access to its essential input made it no longer viable. It was a direct lesson in platform dependency: the ability to build and scale a product is only part of the picture when another company controls the data it needs.

Built with

  • Python
  • YouTube Data API
  • OpenAI API