The scope and depth of human knowledge is so large that validating accuracy of information is humanly impossible for most interesting subjects beyond experts and the very interested.
You're absolutely correct that information needs to be validated, but wrong in singling out Wikipedia. This rabbit hole goes pretty far. I've found false citations from published articles. Conservatively assuming 25 citations per article, the amount you have to validate is 25^(citation depth).
There are 11 citations in the above quote, so here's the homework you're requesting:
Level 0 validation: 11 Articles
Level 1 validation: 275 Articles
Level 2 validation: 6875 Articles
The knee-jerk distrust of Wikipedia is so old that a paper was published in Nature in 2005 showing accuracy close to that of the Encyclopedia Britannica.
https://en.wikipedia.org/wiki/Reliability_of_Wikipedia
Just like any other source, you have to assess it. How many editors, to which degree is it a target for vandalism or astroturfing, how well is the article written, the type of the subject material, etc.
Therefore I would expect britannica-level accuracy on the caffeine page, and that's pretty good for most discussions.