In advancing smart environmental monitoring within IoT frameworks, this paper leverages Wi-Fi channel state information (CSI) for ambient temperature measurement. The study explores the accuracy of temperature predictions and the impact of heat sources on CSI data. Utilizing the ESP32 chipset, we examine its performance in temperature detection under various environmental conditions and the nuanced responses of CSI to thermal variations. Machine learning models emerge as powerful interpreters of CSI data, with a random forest model demonstrating a notable 94.75% accuracy in distinguishing the status of heat sources. Moreover, an LSTM neural network, after extensive tuning and validation, showcases a mean absolute error as low as 0.2416 on focused datasets and 0.5995 on generalized datasets for temperature prediction. These metrics not only underline the efficacy of CSI in environmental monitoring but also highlight the ESP32’s potential as a cost-effective alternative to conventional sensors. This research underscores the transformative implications of Wi-Fi CSI in IoT applications, presenting it as a scalable and economical solution for real-time, smart environmental sensing.