Many studies on phenotypic antimicrobial resistance (AMR) of bacteria from healthy populations are conducted on freeze-stored samples. However, the impact of this practice on phenotypic AMR is not known. We investigated the prevalence of phenotypic AMR in Escherichia coli from chicken (n = 10) and human (n = 11) faecal samples collected from healthy subjects, subject to freeze storage (−20 °C and −80 °C) for 1, 2, 3, and 6 months. We compared counts of E. coli and prevalence of phenotypic resistance against five antimicrobials commonly used in chicken farming (ciprofloxacin, enrofloxacin, doxycycline, gentamicin, and florfenicol) with samples processed within 24 h of collection. Prevalence of phenotypic AMR was estimated by performing differential counts on agar media with and without antimicrobials. At −20 °C, there was a considerable reduction in E. coli counts over time, and this reduction was greater for human samples (−0.630 log10 units per 100 days) compared with chicken samples (−0.178 log10 units per 100 days). For most antimicrobials, AMR prevalence estimates decreased in freeze-stored samples both in humans and chickens over time. Based on these results, we conclude that results on the prevalence of phenotypic AMR on samples from freeze-stored samples are unreliable, and only fresh samples should be used in such studies.