Unless objective measurements are made before and after, there can be a subjective impression that it has made a difference even if it didn't. Without sufficient testing, a "worst case" on a bell curve distribution might be followed by a "best case" on the identical distribution - which sends completely the wrong signal (the impression that the distributions are different, when in fact these are two samples on the same distribution).
Don't blame the user, these are psychological effects that are very hard to counter. Nonetheless, the scientists (who should be aware of the need for proof) should be making the situation clear. Stupidity reigns when discussion takes place only between the ill-educated or misinformed (as is the wont on social media).