For reducing activation of similarity search in a data deduplication system using a processor device in a computing environment, input data is partitioned into data chunks. A determination is made as to whether to apply a similarity search process for an input data chunk based on deduplication results of a previous input data chunk in the input data.