Think around 5% is from visitors, 10-15% from my own experience and the rest just procrastination research.

Started with the cities I know well, and after that adding on countries or cities close by, main focus has been Europe. At one point I tried to use ratebeer's dataset as a starting point, before they closed down, but it was so horribly outdated and irrelevant that it was more work than sourcing manually.

So I basically look for existing blog-ish top-lists for a city, then try to verify the information with search, social media, untappd, etc. Looking for social proof that the venue is operational and relevant.

To keep it updated I have some very rudimentary monthly tasks to ping a venue's website and notify me on things that signal they're closed. I also email myself a list of 10 random venues with all relevant links daily, so I can do a manual 5 min alive check.