Introduction We studied whether: (1) a quality‐focused thyroid fine needle aspiration (FNA) service reduced the non‐diagnostic (ND) rate of FNA; (2) the implementation of thyroid FNA selection criteria resulted in higher proportion of Bethesda V and VI cytology and malignant histopathology; and (3) impact of radiologist's level of experience on ND rates. Method The imaging and pathology computer databases were retrospectively searched for all patients who had thyroid FNAs at our hospitals from July 2004 to December 2016. Results Three hundred thirty‐eight and 609 patients in pre‐thyroid and post‐thyroid service groups, respectively, were evaluated. The ND rate of 36.7% in pre‐thyroid service was significantly higher than post‐thyroid service at 14.6% (p < 0.0001). The rates of Bethesda V and VI cytology in the pre‐thyroid service group did not differ significantly from the post‐thyroid group (p = 0.266; p = 0.069). There was no significant difference in the histopathologic malignancy rates between the two groups (p = 0.531). There was no significant difference in the ND rates of radiologists with varying experience (p = 0.873). Discussion On‐site cytology improved the ND rate of ultrasound‐guided thyroid FNA. Implementation of FNA selection criteria did not alter the rates of Bethesda V and VI cytology and malignant histopathology. The experience of radiologists did not significantly influence ND rates.