Background The causal inference between leisure sedentary behaviour (LSB) and rheumatoid arthritis (RA) is still controversial because of potential residual confounding and reverse causality. Methods The present study used publicly available large‐scale genome‐wide association studies (GWAS) of LSB (television watching, computer use, and driving) and RA to perform a two‐sample Mendelian randomization (MR) study to evaluate the causal effect of LSB on the risk of RA. We detected significant causal associations using the multiplicative random effects‐inverse variance weighted (MRE‐IVW) method, the maximum likelihood, robust adjusted profile scores, the weighted median, MR‐Egger regression, and several complementary sensitivity analyses. Risk factor analysis was also conducted to further investigate potential mediators linking causal inference. Results Increased genetic liability to leisure television watching was significantly associated with a higher risk of RA (MRE‐IVW method; OR = 2.46, 95% CI 1.77–3.41; p = 8.35 × 10−8). MR estimates indicated that prolonged leisure computer use was causally associated with a lower risk of RA (MRE‐IVW method; OR = 0.23, 95% CI 0.12–0.46; p = 2.19 × 10−5). However, we found no evidence for a causal effect of leisure driving on the risk of RA (MRE‐IVW method; OR = 0.59, 95% CI 0.10–3.41; p = 0.557). No pleiotropy was detected by the sensitivity analysis. Conclusions This study supports a causal association between prolonged leisure television watching and an increased risk of RA. Additionally, prolonged computer use might be a protective factor for RA.