The concept that spinal manipulation therapy (SMT) outcomes are optimized when the treatment is aimed at a clinically relevant joint is commonly assumed and central to teaching and clinical use (candidate sites). This systematic review investigated whether clinical effects are superior when this is the case compared to SMT applied elsewhere (non-candidate sites). Eligible study designs were randomized controlled trials that investigated the effect of spinal manipulation applied to candidate versus non-candidate sites for spinal pain. We obtained data from four different databases. Risk of bias was assessed using an adjusted Cochrane risk of bias tool, adding four items for study quality. We extracted between-group differences for any reported outcome or, when not reported, calculated effect sizes from the within-group changes. We compared outcomes for SMT applied at a ‘relevant’ site to SMT applied elsewhere. We prioritized methodologically robust studies when interpreting results. Ten studies, all of acceptable quality, were included that reported 33 between-group differences—five compared treatments within the same spinal region and five at different spinal regions. None of the nine studies with low or moderate risk of bias reported statistically significant between-group differences for any outcome. The tenth study reported a small effect on pain (1.2/10, 95%CI − 1.9 to − 0.5) but had a high risk of bias. None of the nine articles of low or moderate risk of bias and acceptable quality reported that “clinically-relevant” SMT has a superior outcome on any outcome compared to “not clinically-relevant” SMT. This finding contrasts with ideas held in educational programs and clinical practice that emphasize the importance of joint-specific application of SMT.