Should we include a predictor into a prediction model if its P value is >0.05 but it improves AUC and/or model fit?

Dear all,

I am in the process of finalizing my multivariable logistic prediction model. However, I am not sure whether to include some specific predictors. I came across many papers that fit predictors in the final prediction model if their p-values are below 0.05.

However, what should I do if my predictor candidate has a p-value > 0.05 (e.g. 0.06) after being included in the final model but it improves the model fit and AUC? Should I still include it?

All the help is very appreciated!