Growing interest in healthcare has promoted the use of symptom checkers, which are online health applications that provide diagnostic information on users’ health. However, their diagnostic accuracy remains low because the existing symptom checkers rely on manually constructed knowledge models through labor-intensive processes or perform diagnoses based on simple pairwise relationships between diseases and symptoms without considering personal health conditions. In this paper, we propose an intelligent health diagnosis technique that exploits automatically generated ontology and Web-based personal health record services. The proposed technique first automatically generates a human disease diagnosis ontology by exploiting two well-established ontologies for diseases and symptoms: a large-scale medical bibliographic database and an open biomedical repository. When a user enters the symptom-based queries, possible diagnoses are identified by analyzing the user’s queries and their health record data via semantic inferences of the automatically generated ontology. Subsequently, the ranked diagnostic results are provided to the user via ranking methods that consider the user’s symptoms, personal health attributes, and multilevel diagnosis. The proposed technique also provides the user’s diagnostic progress information, which can be used to track or monitor the progress of diseases by considering changes in symptoms over time. The proposed technique was evaluated through a comparison with the existing well-known symptom checkers and other related approaches. The evaluation results show that the proposed technique can feasibly help to improve diagnostic accuracy and deliver appropriate diagnostic information for healthcare action by users.