Background: Community health workers (CHWs), often called "health promotion volunteers" in Japan, are individuals who act as a natural helping resource in the community. Aim: This study tested the efficacy of a CHW-based intervention to improve dietary habits among community-dwelling older people in Japan, using a controlled, crossover design. Method: Seventy-eight people aged 65 to 74 years with poor dietary variety living in four administrative districts in Hikone City (Shiga Prefecture, Japan) were nonrandomly allocated to an immediate-intervened group (IIG; n = 41) or a delayed-intervened group (DIG; n = 37). Participants joined a biweekly, four-session program (120 minutes/session), comprising "CHW drama-style lectures," "group discussion among participants and CHWs," "tasting of dishes," and "take-home practical activities." For the initial 2-month period, the IIG received the intervention and the DIG did not. The groups were crossed over for the subsequent 2-month period. The primary outcome measure was participants' dietary variety score (score range: 0-10). Results: The dietary variety score in the IIG significantly increased in the initial 2-month period compared with the DIG (effect size 1.60 points; 95% confidence interval: 0.75, 2.45). The intervention had a similar effect in the DIG in the subsequent 2-month period. Moreover, an analysis within the IIG showed that the intervention effects persisted for at least 2 months after the intervention. Conclusions: The CHW-based intervention improved dietary habits among older people. Our findings provide evidence that a CHW-based natural helping approach is a possible solution to promote healthy aging in the community.