If the two are close matches on other markers, you should count a multi-step mutation as 1 (infinite-alleles count). However, this method is not usually called genetic distance by hobbyist. Generally, it is simply expressed as matches over markers. For example, a mismatch of two at DYS385 on the 37-marker test is a 36/37 match with a difference of 2 on DYS385. Note that many people will incorrectly express it as a 35/37 match.
FTDNA counts genetic difference in different ways, depending on the application. Although dated and many links do work, this is a good explanation:
At that time, their hybrid method (assumed, when referred to as genetic distance) used the step-wise count for all markers except DYS464 and YCAII. With the new GAP pages, I understand they are transitioning to counting all multi-step markers by the infinite alleles method. According to this page, it had not been implemented:
Multi-step differences are found about 5% of the time on single-copy markers in father-son studies. The chances of 2 or more stepwise mutations in a short time frame is much less, so a multi-step difference is more likely to be a single event, when the two are otherwise close matches on the other markers.
Multi-copy markers have an additional complication. They can also undergo mutations by a second mechanism called recLOH, in which a segment of DNA is deleted and replaced by copying another segment. This results in the duplication of one of the alleles of a multi-copy markers. Whenever, we see a multi-step mutation with duplication on these markers, we usually suspect it is one event.