Photonic sensing devices exploit light–matter interactions to deliver non-invasive, real-time, and highly precise detection of physical, chemical, and biological parameters, underpinning applications in healthcare, environmental monitoring, telecommunications, industrial process control, and structural health monitoring. This review synthesises recent technological and material advances across major photonic sensor classes, including photonic crystal (including photonic crystal fibre sensors, surface plasmon resonance (SPR) sensors, optical fibre-based sensors, and wearable photonic devices, with an emphasis on strategies that improve sensitivity, specificity, flexibility, miniaturisation, multiplexing, and system integration. Using a literature-based comparative analysis of operating principles, device architectures, fabrication routes, and application demonstrations, we highlight how photonic bandgap engineering and defect states enable strong photon confinement (with nanocavity Q-factors reported up to ~38,500), while optofluidic integration supports compact, multifunctional sensing (e.g., 3D photonic crystals in microfluidic channels resolving refractive-index changes of ~6×10^-3). For SPR, prism-coupled (Kretschmann/Otto) and emerging grating-, fibre-, and waveguide-based configurations have been advanced by nanostructured materials and 2D layers; performance gains include gold–graphene hybrids reaching ~350 nm/RIU and architectures targeting high-resolution biosensing (e.g. malaria and HIV DNA hybridisation studies using black phosphorus or MoS2/Si3N4 designs). Optical fibre sensors, including intensity-, interferometric-, and grating-based sensors (FBG), are progressing via specialty fibres (tapered, anti-resonant hollow-core, polymer fibres), composite embedding, and hybrid modalities, enabling distributed, EMI-immune monitoring, and AI/IoT-enabled predictive maintenance. Wearable photonic sensors leverage flexible polymers, hydrogels, and nanomaterials (including MXenes) for skin-conformal monitoring. Machine learning (e.g. CycleGAN) can mitigate motion artefacts without accelerometers while improving energy efficiency by approximately 45%. Key barriers include the need for scalable, uniform, and cost-effective manufacturing; long-term stability, reproducibility, and cross-sensitivity; power management; and, for clinical translation, standardization, data privacy, and security. Overall, converging advances in materials, nanostructures, and machine learning are positioning multifunctional, flexible, and spectrally tunable photonic sensors—including emerging bound-state-in-the-continuum concepts—for impactful deployment in precision healthcare, sustainable environmental sensing, and resilient industrial infrastructure.