Over 17 million people experience a stroke episode annually, with 5.9 million deaths. Stroke is diagnosed by physical tests and neuroimaging which need to be performed quickly to determine if the stroke is caused by ischaemia or haemorrhage. Neuroimaging can reliably confirm bleeding, but many patients with suspected ischaemic stroke (up to 40%) are subsequently confirmed to have alternative pathologies e.g., migraine or seizures (stroke mimics) delaying the transfer of stroke patients to an acute stroke unit for early intervention and treatment. Thus, a simple complimentary blood biomarker test to differentiate stroke patients from non-stroke patients with similar clinical symptoms is essential in prehospital and emergency settings for efficient stroke management and prompt treatment. The current 'Gold Standard' technique for detecting protein biomarkers is complex, time-consuming, and requires automated equipment. In this study, we have developed a proof-of-concept of lanthanide-doped upconversion nanoparticle (UCNP)-based optical biosensor platform for detecting glial fibrillary acidic protein (GFAP), a potential stroke biomarker, in human blood serum. The results show a linear response in photoluminescence quenching of UCNP conjugated GFAP antibody with the increasing concentration of GFAP biomarker in human blood serum. This approach can be used in the ambulance and Emergency Department to quickly diagnose a stroke. In the longer term, such techniques can be integrated into a self-assessment kit to monitor those patients who are at risk after strokes.