Consumers or non-medical professionals are progressively going online to seek health information. Despite the increasing number of health information search online, acquiring the correct and relevant information based on consumer's understanding remains a problem. The information retrieved from the Internet may not fit consumer's understanding because the consumer's familiarity with health topic varies. To improve the accuracy of health information search results, this paper investigates the impact of consumer's familiarity on the search behaviour using language models approach. A user experiment was conducted with 60 participants searching on the topic tasks of dengue fever, diabetes mellitus, and gastroesophageal reflux disease. The participants also rated their familiarity with health task topics on the scale of 1 (not familiar at all) to 4 (familiar). This rating categorized the participants into four familiarity groups (F1, F2, F3, and F4). The data analysis involved the transcription of search task data into the sequence of search activities to identify unique search activity patterns between familiarity groups. The results showed that the familiarity with health topics affected health information search behaviour. There were unique search patterns exhibited by groups of participants with different familiarity. In the query stage, participants with less familiarity issued more modified queries than the participants with higher familiarity. In the decision stage, familiar participants were likely to achieve higher search efficacy than less familiar participants. When locating the potential relevant search result, participants in the higher familiarity groups tended to be more successful than the participants in the lower familiarity groups. By analysing the search behaviour, health information search systems could predict the consumer's familiarity to present more relevant and understandable results.