GolfVotes
The YouTube golf lessons golfers keep telling each other to watch.
Why
YouTube's recommendations are tuned for watch time, not for fixing your slice. Meanwhile golfers in online communities answer each other's swing questions every day with a link to the one lesson that finally helped them. Those answers are scattered across years of threads. GolfVotes collects and counts them.
How it was built
- Read the threads. 1.5 million public comments from golf-instruction communities (2014–2026) were scanned for YouTube links: 29,600 comments.
- Read each comment with an AI model. Is it a real lesson recommendation, not a meme, highlight or gear review? Which problem is it about? Does the golfer say it fixed it?
- Count independent golfers. Each golfer counts once per video. A lesson needs at least 2 different golfers recommending it for a problem.
- Add context. An AI model labels each lesson's type (drill, explainer, live lesson, pro swing …), level and a one-line summary, and builds short fix plans from the top lessons.
- Rank per problem by those recommendations, with extra weight when golfers report it actually fixed their problem.
Promises
Nobody can pay to rank higher. Sponsored content or affiliate links, if they ever appear, are labeled and kept out of the rankings. We don't republish anyone's comments or usernames.
Your data
Liked, history and watch later are stored only in your browser.
Comments
Video pages show short excerpts of public Reddit comments, always linked to the original. If you wrote one and want it removed, open an issue at GitHub with the link and it will be removed in the next update.
Keyboard shortcuts
/ search · h home · k play · l like · w watch later · ? all shortcuts · Esc close