Abstract Background Despite significant success in the fight against malaria over the past two decades, malaria control programmes rely on only two insecticidal methods: indoor residual spraying and insecticidal-treated nets. House improvement (HI) can complement these interventions by reducing human-mosquito contact, thereby reinforcing the gains in disease reduction. This study assessed the implementation fidelity, which is the assessment of how closely an intervention aligns with its intended design, feasibility, and sustainability of community-led HI in southern Malawi. Methods The study, conducted in 22 villages (2730 households), employed a mixed-methods approach. Implementation fidelity was assessed using a modified framework, with longitudinal surveys collecting data on HI coverage indicators. Quantitative analysis, employing descriptive statistics, evaluated the adherence to HI implementation. Qualitative data came from in-depth interviews, key informant interviews, and focus groups involving project beneficiaries and implementers. Qualitative data were analysed using content analysis guided by the implementation fidelity model to explore facilitators, challenges, and factors affecting intervention feasibility. Results The results show that HI was implemented as planned. There was good adherence to the intended community-led HI design; however, the adherence could have been higher but gradually declined over time. In terms of intervention implementation, 74% of houses had attempted to have eaves closed in 2016-17 and 2017-18, compared to 70% in 2018–19. In 2016–17, 42% of houses had all four sides of the eaves closed, compared to 33% in 2018–19. Approximately 72% of houses were screened with gauze wire in 2016-17, compared to 57% in 2018-19. High costs, supply shortages, labour demands, volunteers’ poor living conditions and adverse weather were reported to hinder the ideal HI implementation. Overall, the community described community-led HI as feasible and could be sustained by addressing these socioeconomic and contextual challenges. Conclusion Our study found that although HI was initially implemented as planned, its fidelity declined over time. Using trained volunteers facilitated the fidelity and feasibility of implementing the intervention. A combination of rigorous community education, consistent training, information, education and communication, and intervention modifications may be necessary to address the challenges and enhance the intervention’s fidelity, feasibility, and sustainability.