The simulation of crack-induced failure is of vital importance for the safety and integrity assessment of solids and structures but still challenging. Being the foremost approach, fracture mechanics is plagued by the logical inconsistency in dealing with crack and non-cracked region, and to make matters worse, it has been contradicted by the ongoing experiments that the fracture energy is not a material constant irrelevant to geometry even for the unsophisticated situations. In this paper, based on the recently proposed nonlocal macro-meso-scale consistent damage (NMMD) model, it is demonstrated that the logical consistency of governing equations either on crack-tip or off crack-tip plays a crucial role in the complete solution of cracking problems. To this end, the central themes of fracture mechanics, including the definition of an idealized crack, the separated equilibrium laws and the cracking criteria attached to crack-tip, are firstly revisited. Then, after a concise description of NMMD model with emphasis on the geometrical transmission form micro- to macro-scale and the physical–mechanical translation from geometry to energy, it is theoretically proved that the geometry-based damage zone reproduces idealized crack as a limit when the nonlocal characteristic length, i.e., the influence radius in NMMD model, tends to zero. As a result, the separated equilibrium laws of fracture mechanics can be covered by NMMD model with a unified set of governing equations that does not distinguish between points at crack surface and at non-cracked region. This advantage permits NMMD model to be well-suited to analyze both crack-induced failure problems with and without pre-existing cracks in the absence of additional cracking criteria. In contrast, owing to its macro-meso-scale consistent feature, a new criterion for kinking angle of crack, the maximum stress intensity factor (SIF) of mode-I, can be derived from NMMD model. Some representative numerical examples are also presented to further confirm the above intrinsic link to fracture mechanics. In particular, the superiority of the NMMD model is demonstrated by capturing not only the non-constant fracture parameters (especially the fracture energy) but also the cracking initiation not from the pre-existing crack-tips, which are revealed by experimental results but contradict fracture mechanics. These investigations imply that the NMMD model could provide a potential avenue to understand when and how cracks nucleate and propagate based on a two-scale perspective.