A key metric in tuberculosis epidemiology is the annual risk of infection (ARI), which is usually derived from tuberculin skin test (TST) and interferon-γ release assay (IGRA) prevalence surveys carried out in children. Derivation of the ARI assumes that immunoreactivity is persistent over time; however, reversion of immunoreactivity has long been documented. We used a deterministic, compartmental model of Mycobacterium tuberculosis (Mtb) infection to explore the impact of reversion on ARI estimation using age-specific reversion probabilities for the TST and IGRA. Using empirical data on TST reversion (22.2%/year for persons aged ≤19 years), the true ARI was 2-5 times higher than that estimated from immunoreactivity studies in children aged 8-12 years. Applying empirical reversion probabilities for the IGRA (9.9%/year for youths aged 12-18 years) showed a 1.5- to 2-fold underestimation. ARIs are increasingly underestimated in older populations, due to the cumulative impact of reversion on population reactivity over time. Declines in annual risk did not largely affect the results. Ignoring reversion leads to a stark underestimation of the true ARI in populations and our interpretation of Mtb transmission intensity. In future surveys, researchers should adjust for the reversion probability and its cumulative effect with increasing age to obtain a more accurate reflection of the burden and dynamics of Mtb infection.