The concept of a HATEOAS API is also very simple: the API is defined by a communication protocol, 1 endpoint, and a series of well-defined media types. For a website, the protocol is HTTP, that 1 endpoint is /index.html, and the media types are text/html, application/javascript, image/jpeg, application/json and all of the others.
The purpose of this system is to allow the creation of clients and servers completely independently of each other, and to allow the protocols to evolve independently in subsets of clients and servers without losing interoperability. This is perfectly achieved on the web, to an almost incredible degree. There has never been, at least not in the last decades, a big where, say, Firefox can't correctly display pages severed by Microsoft IIS: every browser really works with every web server, and no browser or server dev even feels a great need to explicitly test against the others.