Gelagen Official Website › RDA Versus AI

Journal

Why Some Rows Get An RDA And Others Only Get An Adequate Intake

By the Gelagen Team Last updated September 30, 2026 All posts · Reviewed by Dr. Jodi Langer, MD

Two different kinds of number hiding behind one percentage sign

A Supplement Facts panel prints one percentage column, and every figure in it is calculated the same way: the declared amount divided by a fixed Daily Value. But the Daily Value itself is not invented by the labeling rule out of nothing. For almost every nutrient on the panel, it traces back to a reference intake published separately by the Food and Nutrition Board, part of the National Academies of Sciences, Engineering, and Medicine, as one of its Dietary Reference Intakes.

That reference intake arrives in the Board’s publications as one of two distinct kinds of figure: a Recommended Dietary Allowance, abbreviated RDA, or an Adequate Intake, abbreviated AI. Both are legitimate, both get used to set Daily Values, and both show up as ordinary rows on an ordinary panel with no visible marking to say which kind a given row is. The difference between them is not about the nutrient’s importance. It is about how the number was calculated, and specifically about how much evidence existed to calculate it with.

What has to be true before an RDA gets set

Before the Board can publish an RDA for a nutrient, it first needs to establish an Estimated Average Requirement — the intake level estimated to meet the needs of half of the healthy people in a given group, based on a specific, measurable indicator of adequacy tracked across a body of dose-response evidence. That indicator might be a blood marker, a functional outcome, or another measurable sign that intake is sufficient, but it has to be something researchers can actually observe changing with dose.

Once that Estimated Average Requirement exists, the RDA is calculated from it directly — typically the average requirement plus roughly two standard deviations, a margin built in specifically so the resulting figure covers the needs of nearly everyone in the group, not just the average person. An RDA, in other words, is not a guess dressed up as a number. It is the output of a defined calculation that only runs once the underlying requirement has been measured with enough precision to support it.

That precision requirement is the bottleneck. Measuring an Estimated Average Requirement means identifying an indicator that changes predictably with intake — a blood level, an enzyme activity, a clinical sign — and then running or assembling enough dose-response evidence to say with confidence where that indicator crosses from insufficient to sufficient for half the group being studied. For a nutrient where no single indicator behaves that predictably, or where the studies needed to map the dose-response relationship have simply not been done at sufficient scale, the calculation cannot run, no matter how well understood the nutrient is in other respects.

What happens when that evidence is not there

Not every nutrient has reached that point. For some, the evidence needed to identify and measure a reliable indicator of adequacy across a dose-response relationship simply does not exist yet — too few controlled studies, no indicator that predicts adequacy consistently, or population data too thin to calculate a defensible average requirement. In those cases, the Board sets an Adequate Intake instead of an RDA.

An Adequate Intake is built differently. Rather than being calculated from a measured requirement, it is based on observed or experimentally determined estimates of the amount of a nutrient that groups of apparently healthy people actually consume — intake levels assumed to be adequate because the people consuming them show no signs of inadequacy. It is a description of what healthy intake looks like in practice, published as a working reference figure, rather than a number derived from a calculated requirement the way an RDA is.

In practice that means an AI is a more conservative kind of figure in one specific sense: it describes intake, not requirement, so it carries a built-in safety margin of its own, on top of whatever margin a comparable RDA would carry. A group of healthy people observed consuming a given amount, with no sign of inadequacy, could in principle need somewhat less than that amount; the AI does not try to isolate the true minimum the way a calculated Estimated Average Requirement does, because the evidence to make that finer distinction is exactly what is missing.

This panel, sorted by which kind of number backs each row

Applying that distinction to Gelagen’s own thirteen rows sorts them cleanly into three groups: nine rows backed by an RDA, three backed by an AI, and one — inositol — backed by neither. The table below uses the same figure each nutrient’s Daily Value is drawn from; several of these reference intakes vary further by sex or life stage in NIH’s full tables, and this table is not a substitute for reading a given nutrient’s own fact sheet in full.

Reference figures are NIH’s published RDAs and AIs, which also serve as the Daily Values Supplement Facts panels calculate percentages against; printed amounts are from the Gelagen label.

Why an Adequate Intake is not a second-class number

It is tempting to read “the evidence was not strong enough for an RDA” as “the nutrient is not well understood” or “the number is less trustworthy.” Neither follows. Biotin and pantothenic acid, both AI-based on this panel, have never shown clear deficiency effects in healthy people eating ordinary diets — and that absence of observed deficiency is itself part of why the dose-response evidence needed to calculate a formal RDA has never accumulated. Nobody has become deficient under study conditions in large enough numbers to measure a precise requirement, which is a different situation from the requirement being unknown or unimportant.

Choline sits in the same AI category for a different reason. A separate page on this website covers choline’s real, demonstrated deficiency effects in controlled research — liver and muscle damage, reversed on repletion — which shows that an AI designation does not mean a nutrient’s function or its deficiency risk is poorly understood. It means the specific evidence needed to calculate an average requirement, as opposed to observing adequate intake in a healthy group, has not yet been assembled to the Board’s standard for an RDA.

Why the system has two tracks instead of one

The two-track structure exists because the Board’s Dietary Reference Intakes are not published once and left alone; they are reviewed and revised nutrient by nutrient as new evidence accumulates, and a nutrient’s designation can move from one track to the other over time. A nutrient sitting on an Adequate Intake today is not permanently assigned there — it is there because, as of the most recent review, the dose-response evidence needed for a calculated Estimated Average Requirement had not yet reached the bar the Board sets for that calculation. If better evidence arrives, the figure can be recalculated and reclassified as an RDA in a future review.

Running two tracks in parallel, rather than waiting for every nutrient to clear the higher evidentiary bar before publishing any reference figure at all, is what lets a panel print a percentage for all thirteen rows instead of only the ones with the most mature research behind them. An AI is, in that sense, the Board’s way of giving consumers and manufacturers a usable working number now, built on the best observational evidence available, rather than leaving a nutrient with no reference figure until a full calculated requirement exists.

The row that gets neither designation

Inositol sits outside this whole framework. No Dietary Reference Intake — not an RDA, not an AI, not a Tolerable Upper Intake Level — has been established for it at all, which is exactly why it carries no percentage on the panel and appears instead as a footnoted amount beneath the main nutrient rows. A missing designation is a different situation from either an RDA or an AI: it reflects a nutrient the Board has not set any reference figure for, rather than one where the reference figure exists but rests on weaker evidence.

What the letters tell you, and what they do not

An RDA-versus-AI designation says something specific and limited: it describes the kind of evidence behind a reference figure, not the nutrient’s importance, not its safety, and not whether a particular product’s printed amount is enough for any one person. It also does not, by itself, say anything about upper safety limits — some AI-based nutrients on this panel, biotin among them, have no established Tolerable Upper Intake Level, while some RDA-based nutrients, vitamin A among them, do. Reference-figure type and upper-limit status are two separate questions, answered separately for each nutrient, and neither one should be inferred from the other.

What the designation is useful for is exactly what this page has done with it: sorting a panel’s thirteen rows by the kind of science standing behind each percentage, so that a reader knows, row by row, whether they are looking at a calculated requirement or an observed-intake estimate.

That sorting habit generalizes past this one panel. Any Supplement Facts label, for any product, can be read the same way: look up whether a given nutrient’s reference figure is published as an RDA or an AI, and treat that as one additional fact about the row — alongside the printed amount and the percentage — rather than as a judgment about the row’s worth. The percentage tells you how the amount compares to the reference. The RDA-or-AI label tells you how confidently that reference itself was established. Both are useful, and neither one substitutes for the other.