Artificial Intelligence (AI) is transforming education by becoming a supportive tool for students and offering multiple possibilities for educators. However, one of the new challenges lies in creating critical minds capable of evaluating both the input of AI prompts and critically interpreting its results, while making sense of the output received. AI brings Data Analytics closer to professionals in all sectors, enabling the adjustment of regression models even for those with basic statistical skills.

Although ChatGPT and DeepSeek allows for the adjustment of linear regression models (MLR) by considering quantitative explanatory variables in an agile manner, the study reveals the necessity for users to have knowledge of modeling theory when introducing qualitative variables (dummies) to avoid statistically correct models that do not accurately represent the reality of the analyzed dataset.

The study was conducted with students from the Bachelor Degree in Business Administration & Management at EAE Business School Barcelona, in the Data Analytics course. Through a transversal sustainability activity, the objective was to use ChatGPT and DeepSeek to find a predictive and explanatory linear regression model that would allow the evaluation of the impact of CO emissions from cars, based on their fuel consumption, considering the fuel type (Gasoline or Diesel). A dataset containing both qualitative and quantitative variables was provided to the students for the study.

The difference between obtaining and interpreting results from correct and incorrect models is analyzed using parametric statistical methods. The focus is on the number of correct models, using ChatGPT and DeepSeek, considering the input prompts and the evaluation and interpretation of results. The tests indicate significant differences, emphasizing better results for students with stronger theoretical knowledge in the Data Analytics course.

Keywords: Learning Analytics, Statistical Thinking, Problem Solving, Linear Models, Data Management, ChatGPT, DeepSeek, AI.