After the absorption of light, the next steps in any artificial photosynthetic system are the generation, transport, and extraction of photoexcited charge carriers. These steps are well understood in solid state systems based on traditional semiconductors, such as Si and III-V absorbers, resulting in quantum efficiencies close to one for optimized devices. For metal oxide-based absorbers, which have attracted much interest in the past decades because of their relatively good (photo)chemical stability and low cost, the reported efficiencies tend to be much lower. This is often attributed to recombination at defects. While these may indeed adversely affect the efficiency, there are several other fundamental loss mechanisms in metal oxides that are not always fully appreciated. For example, certain metal ions exhibit localized d-d transitions that do not result in mobile charge carriers, resulting in a photogeneration yield < 1. Recent work showed that it is possible to quantify the fraction of mobile photoexcited carriers, from which an upper limit of the achievable photocurrent can be derived [1]. Losses can also occur in the form of recombination or carrier localization during transport. The chance of reaching the interface is often expressed as the carrier diffusion length, which in turn depends on the carrier lifetime and its mobility. Convenient contact-free methods to determine these parameters are time-resolved microwave conductivity (TRMC) and time-resolved THz spectroscopy (TRTS) [2]. Here, one of the challenges is to determine whether the decay in photoconductivity is due to a decay in carrier concentration (i.e., recombination), carrier mobility (trapping, polaron formation), or both. We recently developed a general analysis method for determining the diffusion length which is valid for time-dependent mobilities as well as time-dependent lifetimes [3]. Intringuingly, this method revealed a carrier diffusion length of only 15 nm for BiVO4, which is significantly shorter than previously reported values by us and others. The only way this short diffusion length can be reconciled with the high photocurrent densities reported for this material is by assuming field-assisted charge separation. The role of the electric field is, however, poorly understood in the photoelectrochemistry and photocatalysis communities [4]. In a system without externally applied bias, the electric field does not separate the charges; instead, charge separation is driven by the presence of selective contacts. During my presentation I will review these concepts and discuss how a better understanding of loss mechanisms, the role of the electric field, and selective contacts may help us design more efficient metal oxide-based photoelectrodes.