What a health-claims pre-check cannot see

A health-claim pre-check is at its most useful when it is blunt about what it cannot do. The supplement-claim rules (the structure/function framework in 21 CFR 101.93 and the FTC's substantiation standard) regulate claims whose truth lives almost entirely outside the wording — in clinical trials, lab reports, facility-registration records, and the physical label a shopper holds. A tool that reads your sentence can flag risk patterns in that sentence; it cannot confirm the claim. For a brand the cost of confusing those two things is measured in relabel runs, seized inventory, and consent decrees after the claim has already shipped — so the boundary is not a footnote. It is the thing that makes the check worth trusting for what it actually does.

What it catches

Patterns the published rules can be quoted against: a structure/function claim made without the mandatory 101.93(c) disclaimer; a disease-claim verb sitting next to a named disease; an implied "FDA approved" or "FDA certified"; unsubstantiated "clinically proven"/"studies show" cues with no basis named; and absolute or overbroad health wording. These are all properties of the pasted text, so they are checked against the pasted text — deterministically, with quoted evidence, same result every run. The reference framework is snapshot-dated in the engine version rather than claimed to be live-synced, because no wording tool is.

What it cannot catch

• A present disclaimer and a false claim. A label can carry the exact "not intended to diagnose, treat, cure, or prevent any disease" sentence and still assert a structure/function benefit it cannot substantiate. The disclaimer is necessary; it is not sufficient, and its mere presence proves nothing about the evidence.
• The study itself. "Clinically proven on [endpoint]" may or may not rest on a real, adequately-powered trial that measured that endpoint in that dose. No wording check opens your study binder; whether the evidence is competent and reliable is entirely outside the text.
• The nutrient arithmetic. "High in vitamin C" or "good source of fiber" must clear numeric per-serving thresholds defined in the labelling rules. A checker can flag that you used a nutrient-content descriptor; it cannot see your panel of record to confirm the amount qualifies.
• The rendered label. 101.93 asks that a structure/function claim and its disclaimer be presented together and conspicuously. A two-point disclaimer on the back panel under a giant front-of-pack claim may be "present" in the text yet not conspicuous in the real thing a shopper sees. A text scan cannot see font size, layout, or the seal on the front versus the note on the back.
• The regulator's private mind. FDA and FTC decisions, and the 2022 review that touched structure/function enforcement, run on discretion and facts a checker does not model. The rules describe the neighbourhood; the exact materiality and remedy are not a formula.

Why the boundary matters

A tool that claimed your label was "compliant" would be claiming something nobody outside the FDA and FTC can deliver. Saying the limit up front is what lets the check be trusted for what it knows: that this sentence names a disease, that the disclaimer is absent, that a "proven" cue has nothing cited, that the fix is a narrower function statement, an added disclaimer, or a named study — not a dice-roll. The human keeps holding the context the machine cannot see: the trial, the lab, the registration, the market.

The disclaimer is a floor, not a ceiling

One asymmetry surprises brands: adding the 101.93(c) disclaimer fixes a missing-disclaimer gap but does nothing for a disease claim, and that is the point. The disclaimer exists to tell shoppers that a lawful structure/function statement has not been FDA-reviewed; it is not a magic sentence that makes "prevents diabetes" legal, because that sentence is already outside the category the disclaimer governs. A useful pre-check reports both facts separately — "this is a structure/function claim with no disclaimer" (fixable by adding the sentence) and "this names a disease you claim to treat" (not fixable by any sentence, only by removing the claim) — rather than collapsing them into a single "add a disclaimer" suggestion. Getting that distinction right is what stops a brand from papering over a real boundary error with boilerplate and shipping a label that reads as compliant while making an unapproved drug claim. The wording-level check cannot tell you your evidence is strong; it can keep you from confusing a fixable formatting gap with an unforgivable category error.

How to use the boundary productively

A pre-check is the first layer of a health-claim governance stack, not the whole stack. Mechanical wording scan → an evidence pass you run from your own studies and panels of record → a legal pass for jurisdiction and for any borderline structure/function call. Serious brands buy these layers separately; the mistake is treating any one of them as covering the others. Use the check the way the rules invite: catch the disease-boundary and missing-disclaimer errors where they are still a sentence to edit, and route everything that needs real evidence to the people who hold it.

Heuristic pre-check only — not a health-claims certification or legal advice. No clinical-trial, lab, facility-registration, or FDA-approval verification, and no enforcement prediction, is performed. For binding guidance, consult 21 CFR 101.93 and a qualified professional.