Years ago, and expert and AHP creator, Dr. Tomas Saaty answered that question, asserting that AHP is already fuzzy and then, no further fuzzy treatment was necessary
In FAHP, if there is more than one expert involved in the assessment (i.e. group preference), then their judgments (opinions) can be aggregated using weighted averaging method (e.g. establishing weights for experts based on their years of experience). The total of the weights should add up to 1. Then, you multiply each expert's judgment (fuzzy number) by their weight and finally add all resulting judgments fuzzy numbers together in order to get the aggregate one.
You’re right! Tomas Saaty explicitly criticised fuzzifying the AHP method and stated that there is already uncertainty inherent in the nature of the method: i.e. the comparison judgments are already fuzzy because they are allowed to vary over the values of a scale. However, many comparative studies have been made for AHP and FAHP and generally, results indicted that using FAHP leads to better model outcomes than AHP especially when the evaluations made by the decision maker are not much certain.