It needs to be paired with quality experiments, and when that data shows consistent results, those results should be accepted. Asking things like “is the Earth flat” or “does the Covid vaccine kill people” without accepting the wealth of data that’s already out there is of little value.