Words matter: matching language to biological certainty in assisted reproduction

Assisted reproduction depends on concise labels to translate complex gamete and embryo phenotypes into clinically useful information. Yet such labels do not remain purely descriptive. Morphology-based classifications can influence whether an embryo is biopsied, cryopreserved, transferred or discarded; shape patient understanding; define research variables; and become training targets for artificial-intelligence models. The issue is therefore not simply semantic, but how observations acquire inference as they move into decisions, datasets and algorithms. A deviation from an expected phenotype does not necessarily indicate dysfunction, a lower morphological rank does not establish non-viability, and a reference limit is not an individual diagnosis. We argue that terminology in assisted reproduction should be proportional to biological certainty and preserve the distinction between observation, phenotype, ranking, prognosis and demonstrated abnormality. We further propose a pragmatic semantic framework for maintaining those distinctions across clinical communication, research and data-driven applications.