This repository contains the de-identified quantitative dataset and statistical analysis code used to reproduce the findings of a study on interaction with AI companions and psychological well-being. The dataset includes variables such as subjective well-being, chatbot interaction intensity, self-disclosure to chatbots, offline social network size, companionship usage indicators, and demographic controls. It also provides an R Markdown file for multivariate regression and Heckman selection models, alongside a browser extension source code for collecting chat-history donations from Character.AI users, all designed to support research on AI companionship and well-being.