The discrepancy between the widespread use of botulinum neurotoxin (BoNT) in managing headache and the supporting clinical evidence is unprecedented. No substance seems to have inspired more physicians and patients to undertake spirited treatment attempts. Tremendous treatment success in small, uncontrolled clinical trials has been repeatedly reported, but no substance that has been studied to an equal extent has so utterly failed to provide proof of effect in controlled clinical trials. Nevertheless, even though most randomized, controlled clinical trials have not met their defined primary outcome criterion, BoNT is still considered a promising treatment alternative for primary headache disorders. Experimental approaches to the pathophysiologic impact of BoNT on the perception of pain have been equally unsuccessful. Although most studies have been unable to find a direct antinociceptive effect in humans, some researchers continue to seek specific injection sites or injection techniques that may promise more successful results. Others look for a positive effect by narrowing the indications for BoNT to more homogenous symptoms or special patient subgroups. The results of randomized, controlled studies involving a total of 3552 patients indicate that BoNT injection is probably ineffective for patients with migraine and chronic tension-type headache regardless of injection site, dosage, or injection regimen, and there is insufficient evidence to draw a conclusion about its effectiveness for the treatment of chronic daily headache or subforms. The lack of direct experimental or clinical trial evidence that BoNT has a direct antinociceptive effect in humans must be addressed before more trials are conducted, involving even more patients. Additional pathophysiologically oriented research is also needed to unravel the mechanisms of action of BoNT in human pain perception or, alternatively, to bring it all down to the placebo effect.