Do large earthquakes start with a precursory phase of slow slip?

In a recent publication, we showed that a stack of all GPS time series recorded before Magnitude ≥ 7.0 earthquakes suggests that large earthquakes start with a precursory phase of accelerating slow slip (Bletery and Nocquet, 2023). While no peer-reviewed comment or publication has formally contradicted this result, informal discussion has emerged on various platforms. We present here the different elements of discussion and address them through a series of tests. In particular, it has been proposed that correcting GPS time series from network common-mode noise makes the signal vanish. We confirm this result, but we show that this common-mode filtering procedure may inadvertently remove an existing tectonic signal. Moreover, the analysis of past records indicate that the likelihood that common-mode noise produces the signal we observe is well below 1 %. Additionally, we find that the signal is maximum at the location of the impending earthquakes, and for a slip direction (rake angle) close to the one of the upcoming events. The collective outcomes of these tests make very unlikely that the signal solely arises from noise. Even though the results of our tests do not irrefutably demonstrate the existence of a precursory phase of slow slip, they do support its existence. We hope that this study will motivate further work by others to provide a definite answer to the question of the tectonic origin of the observed signal and confirm or refute that large earthquakes start with a precursory phase of slow slip.