Weed management in rainfed rice cropping systems in Madagascar primarily relies on manual weeding. This labour-intensive practice often coincides with peak workloads on the smallholder farms, leading to delayed weeding, one of the major causes of rice yield losses. The aim of this study was to determine the critical period of weed interference (CPWI) in the specific context of low-input rainfed rice-based cropping systems under conventional tillage (CT). We further assessed whether no-tillage with a living mulch of Stylosanthes guianensis (NTLM) could postpone the optimal timing of weeding as cover crops and mulch are often proposed as a promising strategy to reduce weed infestation. Field experiments were conducted during the 2016/17 and 2017/18 cropping seasons. Two controls (weedy or weed-free during the full crop cycle) and six treatments combining two different weeding regimes, increasing weedy period or increasing weed-free period, to three durations (20, 40, and 60 days after sowing) were compared. Rice grain yield was about 2.69 ± 0.32 t ha-1 and 0.04 ± 0.04 t ha-1 in weed-free and weedy controls, respectively. Weed interference throughout the entire rice growing season caused yield losses of up to 99% in CT. In CT, the CPWI started 10 to 18 days after sowing in both years. By contrast, the NTLM system delayed the onset of the CPWI by about 18 days, reducing yield losses to 51% (rice yields of 3.3 ± 0.36 t ha-1 and 1.55 ± 0.51 t ha-1 in weed-free and weedy controls, respectively). Our results highlight the importance of early weeding in rice fields to avoid significant grain yield losses. No-tillage with a living mulch is a promising option, providing additional time to farmers during peak labour periods at the start of the rice growing season. However, further research is needed to confirm these findings and to identify the conditions and support mechanisms that would enable farmers to adopt this type of no-tillage cropping system.