Medical Research: Emerging Concepts and Applications Vol. 2

Laser vision correction has developed over four decades from a single ablative concept into a family of procedures that differ in tissue architecture, optical customisation and biomechanical consequence, and it is now being reframed by the arrival of artificial intelligence in candidate selection, treatment planning and outcome prediction. This critical narrative review examines what the accessible evidence actually supports across that trajectory, and where confident claims outrun the data. Peer-reviewed literature was identified through Europe PMC, Crossref, the Directory of Open Access Journals, Semantic Scholar and Google Scholar, supplemented by citation searching, with a final search date of 23 June 2026. Evidence was appraised for design adequacy, follow-up duration, outcome definition, comparability and independence from commercial sponsorship, and was synthesised thematically rather than procedure by procedure. Three findings recur. First, comparative studies of surface ablation, flap-based keratomileusis and lenticule extraction show broadly convergent efficacy and predictability in low and moderate myopia, so that platform choice rests on secondary considerations rather than on demonstrated superiority of visual outcome. Second, corneal biomechanics, not optical precision, constitutes the principal residual constraint, and thickness-based safety rules remain approximate proxies for a stress response that varies between individuals and between procedures. Third, artificial intelligence in refractive care is at present strongest in narrow, well-specified tasks such as ectasia detection and nomogram optimisation, and weakest in the open-ended judgement that surgical planning requires; reported performance is frequently derived from single-centre, geographically concentrated datasets with limited external validation. The description of contemporary refractive practice as artificial-intelligence-driven therapeutics therefore anticipates a transition that has begun but is not complete. Priorities include prospective multicentre validation of predictive models, standardised biomechanical endpoints, longer follow-up in eyes treated near safety thresholds, and transparent reporting of dataset provenance.