Abstract: Because their great efficiency in producing higher flow rates per unit pressure drawdown, horizontal wells have currently become a popular alternative for the development of hydrocarbon fields around the world. So far, most of the introduced correlations to estimate the productivity index for these wells have shown certain differences among their results. This does not allow us to properly establish which one of them provides the closest value to the actual one, since there is no evidence of a trustable enough reference point. Throughout the years, several investigations for the determination of horizontal-well productivity index have been carried out. These researches have been focused on the determination of steady-state solutions for the above-mentioned parameter, therefore, a diverse number of correlations have been introduced. This study proposes an improved steady-state correlation to calculate productivity index for horizontal wells and evaluates the most commonly used existing correlations to estimate this parameter by using numerical simulation. Besides that, a sensitivity analysis on the influence of the variation of each variable in the existing and proposed models was carried out. The analysis was conducted by generating a synthetic drawdown test by means of a commercial reservoir simulator. Using the pressure derivative curve, a time range where steady-state behavior takes place was defined. Then a simulation was performed with the purpose of determining the pressure distribution in the reservoir within that range of time. This allows us to estimate the horizontal-well productivity index for any drainage radius. There is no way to establish what correlation is the most accurate since the productivity index of a horizontal well cannot be physically measured. Therefore, we conducted a sensitivity analysis considering the variation of the most relevant parameters and compared to the results from a simulator. More than 500 simulation runs were performed to estimate the results obtained by the improved correlation introduced in this work and the existing ones. Several plots of productivity index versus each one of the model variables were constructed for comparison purposes leading to obtain an application range for the existing correlations; among these, we observed that Joshi�s correlation matches well with the simulated results. However, the proposed correlation provides much better results than those provided by Joshi�s within a very wide range of variation of the parameters involved in the different correlations. We did not provide any value of the deviation error since this normally increases as the the studied parameter increases. However, comparative plots speak by themselves.