Owing to the extensive building projects that are currently underway in China, evaluation of the embodied emissions in the building construction phase is crucial for reducing the carbon footprint. Previous studies have focused on the quantity of emissions, whereas the present study focuses on the issue of uncertainty in building emission assessment. In this context, a semi-quantitative approach was adopted, and the probabilistic distributions of the quantities and emissions of building materials and energy were assessed based on data quality indicators. Further, a case study was conducted to compare the deterministic and stochastic emissions. The results showed that the sample mean of the stochastic results (5891.97 tCO2e) was consistent with that of the conventional method, while the relevant standard deviation was estimated as 248.90 tCO2e owing to the uncertainty of input parameters. In addition, scenario analyses were conducted, including the system boundary, potential reduction of material consumption and emission, and adoption of local production and low-carbon energy to quantify the scenario uncertainty, and the transformation coefficients and temporal correlation to quantify the model uncertainty. The relevant analyses revealed the key factors (e.g. system boundary, steel, concrete, and masonry works, local production, and applicable period of the data) in reducing the embodied emissions and corresponding uncertainties. Overall, the present study can facilitate comprehensive assessment of the uncertainties in building embodied emissions, thereby contributing to low-carbon policy-making in the building industry.