The discovery of Earth-like planets is a major focus of current planetology research and faces a significant technological challenge. Indeed, when it comes to detecting planets as small and cold as the Earth, the cost of observation time is massive. Understanding in what type of systems Earth-like planets (ELPs) form and how to identify them is crucial for preparing future missions such as PLATO, LIFE, or others. Theoretical models suggest that ELPs predominantly form within a certain type of system architecture. Therefore, the presence or absence of ELPs could be inferred from the arrangement of other planets within the same system. This study aims to identify the profile of a typical system that harbours an ELP by investigating the architecture of systems and the properties of their innermost detectable planets. Here, we introduce a novel method for determining the architecture of planetary systems and categorising them into four distinct classes. We then conduct a statistical study to identify the most favourable arrangements for the presence of an ELP. Using three populations of synthetic planetary systems generated using the Bern model around three different types of stars, we studied the `theoretical' architecture (the architecture of a complete planetary system) and the `biased' architecture (the architecture of a system in which only detectable planets are taken into account after applying an observation bias) of the synthetic systems. To describe a typical system hosting an ELP, we initially examined the distribution of ELPs across different categories of architectures, highlighting the strong link between planetary system architecture and the presence of an ELP. A more detailed analysis was then conducted, linking the biased architecture of a system with the physical properties of its innermost observable planet to establish the most favourable conditions for the presence or absence of an ELP in a system. First, using synthetic systems, we successfully reproduce the distribution of observed multi-planet systems within the five different architectural classes. This demonstrates the relevance, at the system level, of populations of the synthetic systems derived from the Bern model and the observational bias applied. Secondly, the biased architectures (with observation bias) correspond for the most part to the theoretical architectures (without bias) of the same system. Finally, the biased architecture of a system, studied in conjunction with the mass, radius, and period of the innermost detectable planet, appears to correlate with the presence or absence of an ELP in the same system. We conclude that the detections of ELPs can be predicted thanks to the already known properties of their systems, and we present a list of the properties of the systems most likely to host such a planet.