The purpose of this study is to present a Fuzzy Automatic New Topic Identification System (FANTIS) to estimate topic changes in search engine transaction logs. Sample datasets of two search engines are used for the illustration of the approach. A two-input, one-output and three-rule fuzzy system is designed using the general topic continuation and shift distribution information in the datasets. The system is then used for automatic new topic identification. Our findings show that FANTIS can successfully be used as an automatic new topic identification tool. Compared to the other studies of automatic new topic identification, in addition to its satisfactory performance, FANTIS stands out as a flexible and simple approach since (i) it can be easily modified and (ii) it does not require a formal training phase. The proposed approach contributes to the solution of the automatic new topic identification, which constitutes one of the main problems in information retrieval that need to be solved for achieving the goal of personalized search engines to yield more efficient search sessions.