During Run 2, the LHC delivered instantaneous luminosities of approximately $10^{34}\ \textrm{cm}^{-2} \textrm{s}^{-1}$ at $\sqrt{s}=13$ TeV in $pp$ collisions. At such high instantanenous luminosities, measuring the decay rate of $Z\to\ell\ell$ provides a powerful tool to monitor the luminosity recorded by ATLAS over time periods as short as 60s.These proceedings present an overview of the method, outlining the event selection, data-driven efficiency determination and corrections derived from simulation, as well as showcasing the robustness of the final results. The absolute luminosities obtained independently in both the $Z\to ee$ and $Z\to\mu\mu$ channels agree to within approximately 1\%, with an excellent time stability of around 0.5\%, a non-trivial result considering both channels have their own distinct chain of corrections. The difference between the normalised $Z$-counting luminosity and ATLAS baseline luminosity, where the $Z$-counting luminosity is normalised to the same integrated luminosity as the ATLAS baseline measurement over the entire data-taking period, is evaluated per LHC fill and found to typically be within 0.5\%.