The social media adoption rate and business models make them an efficient environment for influencing users with disinformation content. In turn, the process of automated or semi-automated disinformation detection is challenging for many reasons. Numerous research methods and applications are published in this area, but they suffer from significant limitations related to their functionality or detection accuracy. These circumstances are the motivation for the presented work, which contributes with: a novel multi-criteria disinformation detection method implemented in the XScanner prototype application dedicated to the X service and a series of experiments designed to evaluate the properties of the XScanner method. XScanner application has the ability to operate on online data, to analyze the content, metadata, and interactions, and to produce fine-grained results for the analyst. XScanner LDA efficiency of 91% and F1-score of 0.87 show that the model maintains a balanced performance, effectively minimizing both false positives and false negatives. Through statistical analysis, the findings confirm that the identified criteria as well as XScanner application can be both effectively used to support a fact checker, community manager or security manager in the process of disinformation detection and labeling.