Objective: The 2014 Bethesda System diagnostic criteria for atypical glandular cells (AGC) aids in classification of atypical cells in cervical cytology. There is limited literature regarding reproducibility and interobserver variability in the application of the 2014 AGC criteria. Our aim is to assess the interobserver variability of AGC with a focus on how diagnostic categories link with guideline-driven management. Study Design: Three observers re-reviewed 51 previously diagnosed AGC Papanicolaou tests. The diagnoses were categorized as follows: (1) according to guideline-specified management, and (2) as glandular vs. squamous lesions. The κ statistic was used to evaluate interobserver agreement. Results: The interobserver variability per guideline management by weighted 2-observer κ ranged from 0.009 to 0.530, with half of the interobserver pairings meeting the threshold for at least fair-moderate agreement. For categorization as glandular, squamous, or both, unweighted κ yielded at best fair interobserver agreement (κ = 0.250) in 1 pairing, with low κ scores in the remainder of reviewer pairs (range 0.015–0.152). Conclusions: There is significant interobserver variability in the diagnosis of AGC. The AGC cases when divided by clinical management had fair-moderate interobserver agreement, suggesting that diagnostic variability likely has a real effect on patient care. This diagnostic uncertainty should be understood by cytologists and clinicians.