The reviewer of my original proposal recommended the use of the t-test. While the t-test is indeed very handy, I believe that the very large n (sample size) allows us to forgo the use of the t-test. Moreover, the t-test is an updated version of the z-test for tests of low n. Thus, the z-test should still work. However, a Chi Square test should be the better fit nonetheless due to the categorical nature of the data that we have created - user age group being a binary.
We did perform such a Chi Square test using different years as extra categorical variables with the idea of seeing if this proportion held up over different years. However, the test resulted in a very high p-value.