The claim is the complete communicated message
Regulators evaluate more than one isolated sentence. Product name, surrounding images, testimonials, comparisons, search keywords, page structure and calls to action can combine into an express or implied message. Replacing a disease term with wellness language does not solve the problem when the overall presentation continues to promise treatment or prescription-equivalent performance.
A rigorous system therefore represents intended use as a structured object. It records the proposed statement, audience, placement, linked evidence, jurisdiction and neighboring content. The review question becomes reproducible: what would a reasonable reader understand, and does the evidence support that understanding at the required level?
Structure/function is a defined category
FDA describes structure/function claims as statements about the role of a nutrient or dietary ingredient in affecting the normal structure or function of the human body. General well-being statements can also exist within the dietary-supplement framework. The exact ingredient must still qualify for the category, and the statement must be truthful and non-misleading.
This framework differs from a claim to diagnose, treat, cure or prevent disease. The boundary is not created by a disclaimer alone. Wording, context and scientific support work together. Ingredient eligibility, manufacturing controls, label requirements and required notifications are separate obligations that remain even when the sentence itself fits a structure/function form.
Substantiation must match specificity
FTC guidance focuses on competent and reliable scientific evidence and on the fit between evidence and the advertised message. A study on a different ingredient, dose, preparation, population or endpoint may provide useful background while failing to substantiate a specific finished-product claim. Mechanistic plausibility is not the same evidence type as a controlled human outcome.
Specificity raises the evidentiary burden because it narrows what the statement asks the evidence to establish. Comparative, quantitative, rapid-onset, safety and serious-condition claims each introduce additional propositions. An evidence graph should split these propositions rather than attaching one citation to a sentence that silently contains several different promises.
Compile language from evidence, not evidence from language
The NOVA-4 publishing architecture starts with evidence objects carrying source class and scope. A general-mechanism source can support an educational explanation. An internal design object can support a statement about what the program intends to measure. Neither can be promoted automatically into a finished-product human-effect statement.
This direction of compilation scales cleanly. New languages, pages and partner materials inherit the same typed claim object instead of improvising translations. When a stronger evidence object is added, the permitted output can expand deliberately. The system increases distribution while preserving one evidentiary meaning across every surface.