Bitcoin has become an increasingly common investment instrument despite its high volatility. There is an interest within the financial market in seeking better models for predicting Bitcoin prices. The existing models range from statistical models to machine learning algorithms. Among these models, many use only endogenous variables, while others incorporate some exogenous variables. The aim of our study is to employ a Gated Recurrent Unit (GRU), a variation of a Recurrent Neural Network(RNN), combining both endogenous and exogenous variables, to conduct a series of simulations in order to understand which of the exogenous variables has better accuracy in forecasting Bitcoin's price. For this study, we selected three different exogenous variables, named Google Trends, CBOE Gold ETF Volatility Index (GVZ), and Office of Financial Research - Financial Stress Index (OFR FSI). We used Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) metrics to evaluate the performance of the simulations. The results showed an improvement in accuracy when using the three exogenous variables together. Although there is a difference, it is relatively small, and the model performed worse in the simulations when using only one exogenous variable. Therefore, exogenous variables are a new venue that can and should be further explored, as suggested in our future work considerations.