Social media has become a useful instrument and forum for expressing worries about various difficulties and day-to-day concerns. The pertinent postings containing people’s complaints about water quality as an additional source of information can be automatically acquired/retrieved and analyzed using natural language processing and machine learning approaches. In this paper, we search social media for a water quality analysis and propose a scalable messaging system for quality-related issues to the subscribers. We classify the WaterQualityTweets dataset, our newly collected collection, in two phases. In the first phase, tweets are classified into two classes (water quality-related or not). In the second phase, water quality-related issues are classified into four classes (color, illness, odor/taste, and unusual state). The best performance results are BERT and CNN, respectively, for binary and multi-class classification. Also, these issues are sent to different subscribers via a topic-based system with their location and timing information. Depending on the topics that online users are interested in, some information spreads faster than others. In our dataset, we also predict the information diffusion to understand water quality issues’ spreading. The time and effort required for manual comments obtained through crowd-sourcing techniques will significantly decline as a result of this automatic analysis of water quality issues.