What is the co-watch graph?

The co-watch graph is a mechanism used by platforms like YouTube to improve video recommendations. It functions by tracking and analyzing which videos are frequently watched together by users. For example, if a user watches video A and then video B, and many others do the same, a link is formed between these videos. Consequently, if a new user watches video A, video B might be recommended to them based on this co-watch data.

This method is more computationally feasible than running massive neural networks, as it narrows down the pool of videos using these co-watch links before applying machine learning algorithms to make final recommendations. This process can significantly impact whether a well-made video reaches its intended audience, especially in niche topics like mathematics, by ensuring it is included in the initial pool of recommended videos 1.