Table 2.

Risk of bias summary for included studies (Cochrane Risk of Bias [RoB] 2.0)^a.

Study	Randomization process	Deviations from intended interventions	Missing outcome data	Measurement of the outcome	Selection of the reported result	Overall bias
Lui et al, 2024 [18]	Low	Low	Low	Low	Low	Low
Wang et al, 2020 [19]	Low	Low	Low	Low	Low	Low
Yamaguchi et al, 2023 [20]	Low	Low	Low	Low	Low	Low
Nakashima et al, 2023 [21]	Low	Low	Low	Low	Low	Some concerns
Glissen Brown et al, 2021 [22]	Low	Low	Low	Low	Low	Some concerns
Ahmad et al, 2023 [23]	Low	Low	Low	Low	Low	Low
Wang et al, 2019 [24]	Low	Low	Low	Low	Low	Low
Shen et al, 2021 [25]	High	Low	Some concerns	High	Some concerns	High
Yao et al, 2022 [26]	Low	Low	Low	Low	Low	Low
Liu et al, 2020 [27]	Low	Low	Low	High	Low	High
Xu et al, 2021 [28]	Low	Some concerns	High	High	Low	High
Liu et al, 2019 [29]	Some concerns	Low	Low	High	Some concerns	High
Gong et al, 2020 [30]	Low	Low	Low	Low	Low	Low

^aThe “Deviations from intended interventions” domain was rated as high risk in most studies due to the inherent inability to blind endoscopists and participants to the AI intervention, which is an intrinsic limitation of AI-assisted colonoscopy trials rather than poor study conduct. The “Measurement of the outcome” domain was predominantly low risk as polyp detection counts were objectively measured and histopathologically confirmed. Overall risk of bias was determined per RoB 2.0 guidance: “Low” if all domains were low risk; “Some concerns” if at least 1 domain had some concerns but no domain was high risk; “High” if any domain was high risk or if multiple domains had some concerns that substantially lowered confidence in the result.