The question is natural but the answer is highly individual. For what it's worth, I'll comment from my own perspective as a now-retired reviewer who produced almost 500 reviews over many decades. In its early years the print version of Mathematical Reviews was still relatively slim but covered the most widely circulated journals and had as reviewers most of the active mathematicians of the time. It had the great advantage of bringing together both a print database of current literature and (often) helpful commentaries by specialists. For many decades the authors were compensated only by receiving free papers (sometimes books).
As mathematics and its offshoots proliferated in the 1960s and later, it became impossible for most people to skim all reviews. But the evolving classification scheme helped, even though it could never meet all needs. Until the Internet (and arXiv) developed far enough, the reviews and database played a mostly constructive role in communication. But managing the flow of papers and editing the submitted reviews required a lot of expensive professionals, as it still does. Some reviews were of course eccentric, such as one which simply quoted verbatim half of a two page note from a widely circulated Springer journal.
For me personally it was a way to keep in touch with a wider range of interesting mathematics than I actually worked on at the time. But to do the reviewing task well is time-consuming, since I always felt the need to delve into the related literature. (At least once I discovered an earlier proof in a slightly offbeat journal of a theorem published anew in a mainstream journal by an author who hadn't been aware of the earlier proof.) Sometimes you get correspondence (even arguments) from an author whose work you have reviewed. All very interesting, but optional activity like refereeing.
By now MathSciNet functions mainly as an excellent database, still very expensive to maintain, and reviewers are given AMS credits for their use. Fewer papers get full reviews, which is usually the right decision but not always. People rely more for up-do-date stuff on other Internet sources, but the organized and flexibly searchable database is worth the cost for those institutions which can afford it. (Not all can.)