Ultra-processed food (UPF) consumption, captured by the Nova classification, is robustly associated with adverse health outcomes when considered at the level of the overall diet, and this evidence increasingly informs dietary guidance. Translating this evidence into advice about individual foods is far less straightforward. Nova classifies foods by the degree and purpose of industrial processing rather than by nutritional composition, and its fourth group spans products of markedly different nutritional character, from sugar-sweetened beverages and confectionery to fortified wholegrain cereals and wholemeal breads. Group 4 status is therefore a poor discriminator of products within the group. This is a narrower claim than arguing that processing is unimportant. This opinion paper argues that clinical advice is most useful when matched to the level at which the evidence is strongest, and proposes an evidence-proportionate framework that links the level of evidence to the inference that can reasonably be drawn from it and to the corresponding level of dietary advice. The framework is an operationalisation of existing evidence rather than a new empirical claim, and it specifies the additional information required before acting at each level, including a structured assessment for decisions about a single Group 4 food.