Indeed, the best type of knowledge comes from primary sources and original research. But those also produce an awful lot of not-knowledge.

Wikipedia's approach to sifting the knowledge from the not-knowledge is to prefer reliable secondary sources, i.e. sources deemed capable of telling the difference, mainly because they have a reputation for good editorial control. It's far from an ideal touchstone; but relying on "experts" is worse, because who's an expert? You need experts to identify experts, which is circular.

"Reliable secondary sources" doesn't amount to hearsay.