It still commits the same fundamental mistake made by any other popular sources that covers supplements: it doesn't really give study quality adequate attention. Instead, it just accepts all published findings as probably being valid. What with the replicability rate for medical research papers being somewhere around 20%, that's an approach that will lead you astray 4 times out of 5.
As a concrete example, the support they give for the first item in the table of results, lipid peroxidation, is three papers with n=37, n=16 and n=9. All of them throw some flags that make me suspect p-hacking may have occurred, with the n=37 one being the most worrisome in that department.
