Objective To explore the causal relationship between heart failure and memory impairment using a two-sample two-way Mendelian randomization (MR) approach. Methods The data for this study was sourced from the genome-wide association studies database and analyzed using a two-sample two-way MR method. In the forward study, the exposure factor was heart failure, and the outcome was memory loss. In the reverse study, the exposure factor was memory loss, and the outcome was heart failure. The instrumental variables were selected, and the causal relationship between heart failure and memory impairment was mainly analyzed using the inverse variance weighted method (IVW). At the same time, MR Egger regression, weighted mode (WM), weighted median estimator (WME), and simple mode (SM) were used to supplement the IVW analysis results. Cochran’s Q test was used to assess statistical heterogeneity among SNPs, the intercept term of MR Egger regression was used to evaluate the presence of horizontal pleiotropy among single nucleotide polymorphisms (SNPs), and the leave one method was used to evaluate the impact of individual SNPs on IVW analysis results. To avoid reverse causality, memory loss was identified as the exposure factor and heart failure as the outcome event, and reverse MR analysis was conducted. Results A total of 9 SNPs strongly associated with heart failure. The IVW analysis results [β=−0.81, odds ratio (OR)=0.45, 95% confidence interval (CI) (0.27, 0.73), P=0.0015] showed a negative causal relationship between heart failure and memory impairment. The results of weighted median estimator [β=−0.92, OR=0.40, 95% CI (0.21, 0.77), P=0.0059] and weighted mode [β=−1.16, OR=0.31, 95% CI (0.12, 0.83), P=0.047] showed a negative causal relationship between heart failure and memory impairment. Although MR Egger regression [β=−1.19, OR=0.30, 95% CI (0.08, 1.21), P=0.14] and simple mode [β=−0.32, OR=0.72, 95%CI (0.26, 2.01), P=0.55] analysis showed that heart failure does not increase the risk of memory loss. Cochran’s Q test and prompt yielded robust and non-heterogeneous results (P=0.890), while the intercept assessment of MR Egger regression indicated stable results without horizontal pleiotropy (P=0.578). Reverse MR analysis revealed that there is no causal relationship between exposure factor memory impairment and outcome heart failure [β=3.71×10−3, OR=1.00, 95%CI (0.96, 1.03), P=0.85]. Conclusions There is a negative correlation between genetic prediction of heart failure and the risk of memory loss, which is inconsistent with clinical observations and previous research conclusions. Observational associations may be overestimated or misled, which is also one of the core values of MR research.