Room temperature (RT) point-contact (PC) single electron transistors (SETs) in silicon (Si) allow control of charge on a nanoscale quantum dot (QD) at the one electron level. These devices are of great interest for investigations of few particle physics and ‘beyond CMOS’ nanoelectronics applications. However, RT operation requires ultra-small sub-10 nm QDs, such that the single electron charging energy EC ≫ kBT = 26 meV, the thermal energy at RT, raising major nanofabrication challenges. This work considers the nanofabrication, electrical measurement, and non-invasive optical characterisation of ∼10 nm scale QD SETs. Devices were defined in silicon-on-insulator material with an n-type doped (1020 cm−3) ∼40 nm top Si layer using electron beam lithography and plasma etching. ‘Geometric oxidation’ of the PC region isolated dopant atom QDs. An alternative hybrid SET/Field Effect Transistor was also developed, with partially oxidised PCs. Fabrication process calibration greatly improved device success percentage to ∼37%, well above previous studies with ∼10%. Single electron simulation using an equivalent circuit model allowed extraction of QD parameters. Although fabrication process calibration reduces device failure, significant problems can persist at the ∼10 nm scales and structural inspection is critical. Conventional inspection methods, e.g., scanning electron microscopy, offer few-nm resolution but may be invasive and induce damage. Optical dark-field microscopy enables non-invasive, sub-wavelength investigation of nanostructures. Previously, this method was used to investigate ∼100 nm scale nanoelectromechanical systems and isolated, ∼10 nm scale nanoparticles. Here, we show for the first time that when applied to complete ∼10 nm scale SET structures, fabrication failure can be identified. Optical simulation, based on rigorous modal diffraction theory, was optimised to obtain a speed-up of ∼600× compared to previous reports. This enabled extraction of diffraction peak signatures for typical nanofabrication failures. Hence, comparison to experimental images created a method for non-invasive identification of successfully fabricated devices.