A major societal issue of disposal and environmental pollution is raised by the enormous and fast-growing production of single-use polyethylene terephthalate (PET) bottles, especially in developing countries. To contribute to the problem solution, an original route to recycle PET in the form of value-added environmentally friendly thermoplastic composites with banana fibres (Musa acuminata) has been developed at the laboratory scale. Banana fibres are a so far undervalued by-product of banana crops with great potential as polymer reinforcement. The melt-processing constraints of commercial PET, including used bottles, being incompatible with the thermal stability limits use of natural fibres; PET has been modified with bio-sourced reactants to produce co-polymers with moderate processing temperatures below 200 °C. First, commercial PET were partially glycolyzed with 1.3-propanediol to produce co-oligomers of about 20 repeating units, which were next chain extended with succinic anhydride and post-treated in a very unusual "soft solid state" process at temperatures in the vicinity of the melting point to generate co-polymers with excellent ductility. The molar mass build-up reaction is dominated by esterification of the chain ends and benefits from the addition of succinic anhydride to rebalance the acid-to-hydroxyl end-group ratio. Infra-red spectroscopy and intrinsic viscosity were extensively used to quantify the concentration of chain ends and the average molar mass of the co-polymers at all stages of the process. The best co-polymers are crystallisable, though at slow kinetics, with a Tg of 48 °C and a melting point strongly dependent upon thermal history. The composites show high stiffness (4.8 GPa at 20% fibres), consistent with the excellent dispersion of the fibres and a very high interfacial cohesion. The strong adhesion can be tentatively explained by covalent bonding involving unreacted succinic anhydride in excess during solid stating. A first approach to quantify the sustainable benefits of this PET recycling route, based on a rational eco-selection method, gives promising results since the composites come close to low-end wood materials in terms of the stiffness/embodied energy balance. Moreover, this approach can easily be extended to many other natural fibres. The present study is limited to a proof of concept at the laboratory scale but is encouraging enough to warrant a follow-up study toward scale-up and application development.