Abstract Background Perceived disease risk may reflect actual risk indicators and/or motivation to change lifestyle. Yet, few longitudinal studies have assessed how perceived risk relates to risk indicators among different disease risk groups. We examined in a 5-year follow-up, whether perceived risks of diabetes and cardiovascular disease predicted physical activity, body mass index (BMI kg/m²), and blood glucose level, or the reverse. We examined further whether perceived risk, self-efficacy, and outcome beliefs together predicted changes in these risk indicators. Method Participants were high diabetes risk participants (N = 432) and low/moderate-risk participants (N = 477) from the national FINRISK 2002 study who were followed up in 2007. Both study phases included questionnaires and health examinations with individual feedback letters. Data were analyzed using gender- and age-adjusted structural equation models. Results In cross-lagged autoregressive models, perceived risks were not found to predict 5-year changes in physical activity, BMI, or 2-h glucose. In contrast, higher BMI and 2-h glucose predicted 5-year increases in perceived risks (β-values 0.07–0.15, P-values < 0.001–0.138). These associations were similar among high- and low/moderate-risk samples. In further structural equation models, higher self-efficacy predicted increased physical activity among both samples (β-values 0.10–0.16, P-values 0.005–0.034). Higher outcome beliefs predicted lower BMI among the low/moderate-risk sample (β-values − 0.04 to − 0.05, P-values 0.008–0.011). Conclusion Perceived risk of chronic disease rather follows risk indicators than predicts long-term lifestyle changes. To promote sustained lifestyle changes, future intervention studies need to examine the best ways to combine risk feedback with efficient behavior change techniques.