The stepping-forward affordance perception test (SF-APT) fills an important gap within the screening of falls risk factors by considering the perception of affordances. The test showed to be a valid instrument for community-dwelling older adults falls risk assessment. The present study aimed to distinguish and test the key outcomes of the SF-APT usable for falls risk assessment in community-dwelling older adults to determine the respective cut-offs. This cross-sectional study enrolled 347 participants (73.1 ± 6.2 years; non-fallers: 57.9%; fallers: 42.1%; recurrent-fallers: 17.9%). Falls occurrence and SF-APT outcomes were assessed. Analyses were performed using multivariate binary logistic regression analysis and receiver operating characteristic (ROC). The area under the ROC curve was computed (AUC) for each built model explaining falling or recurrent falling. Results distinguished the Estimated stepping-forward, and Absolute-error in interaction with Error-tendency as the SF-APT key outcomes for falls risk assessment [AUCfalling: 0.665 (CI 95%: 0.608-0.723); AUCfalling recurrently: 0.728 (CI 95%: 0.655-0.797)]. Computed cut-offs' analysis showed that (i) a boundary stepping-forward estimation >58 cm plus an underestimation bias >5 cm (>42nd percentile) avoid older adults to be recurrent-fallers, and (ii) a boundary stepping-forward estimation >62 cm plus an underestimation bias >6 cm (>54th percentile) avoid older adults to be fallers. In conclusion, results suggest that SF-APT is a valuable tool for falls risk assessment in community-dwelling older adults. Interventions targeting the prevention of falls should consider the above key outcomes and the respective cut-offs as alert red-flags.