Sweeteners in Food: Evidence, Exposure and Safety Review

Evidence review updated October 8, 2026. Sweeteners differ in chemistry, intended use and evidence. A product label, a laboratory finding and a clinical outcome answer different questions.

Sweetness Decoded | What Sweetener Labels Leave Unanswered

One yogurt says "zero sugar." Another says "keto friendly." A third promises less sweetness without giving up taste. The front panel makes the decision look simple. Turn the package around and the questions multiply.

Which sweetener is present? What evidence applies to it? How much is in one serving? Is the claim about calories, weight control, blood glucose or long-term safety? A sugar claim does not answer all four.

That is the useful starting point for this article: read the ingredient, the evidence and its limits together. Neither a reassuring label nor a frightening headline can do that work alone.

What the ingredient list can tell you

The FDA explains that sweeteners appear in many foods and beverages and can be identified by their names in the ingredient list. Some uses require food-additive approval; others use the generally recognized as safe, or GRAS, pathway. These are different regulatory routes, both tied to conditions of use.

An ingredient name is a starting point, not a complete exposure measurement. A shopper should not assume that a package provides enough information to calculate the intake of every sweetener or to predict an individual health outcome. The product, serving size, amount and pattern of consumption still matter.

What WHO actually said about weight control

In 2023, WHO issued a conditional recommendation against using non-sugar sweeteners for long-term weight control or prevention of noncommunicable disease. WHO noted that observational links with disease can be affected by confounding, including differences between people who choose these products and those who do not.

The scope matters. The recommendation excludes people with pre-existing diabetes. It also excludes low-calorie sugars and sugar alcohols, or polyols. Erythritol and xylitol should therefore not be presented as ingredients covered by that recommendation. The guideline is not a toxicological verdict that all sweeteners are unsafe.

A person managing diabetes needs advice suited to their own care plan. A general population guideline is not a substitute for that conversation.

Aspartame: hazard is not the same as exposure risk

In July 2023, IARC classified aspartame as possibly carcinogenic to humans, Group 2B, citing limited human evidence. At the same time, JECFA retained its acceptable daily intake of 0 to 40 milligrams per kilogram of body weight. The hazard classification concerns the strength of evidence that an agent can cause cancer; it does not calculate the risk at a specific dietary exposure.

The FDA states that it does not have safety concerns about aspartame under approved conditions of use. Its acceptable daily intake is 50 milligrams per kilogram of body weight. That U.S. value and the JECFA value come from different assessments; they should not be blended into one universal number.

There is an important specific exception: people with phenylketonuria, or PKU, must avoid or restrict aspartame because it contains phenylalanine. That is a concrete individual consideration, not a reason to assign the same warning to every consumer.

Erythritol: a signal worth investigating, with limits

A 2023 Nature Medicine study found an association between higher circulating erythritol and later major cardiovascular events in cohorts undergoing cardiac evaluation. It also included laboratory and animal experiments and an ingestion study in eight volunteers. Erythritol can be produced within the body, so a blood measurement cannot automatically be treated as a record of dietary consumption.

A 2024 study examined 20 healthy volunteers: 10 consumed 30 grams of erythritol and 10 consumed 30 grams of glucose. The erythritol group showed increased platelet responsiveness in the measured short-term tests. These are biological endpoints, not evidence that the participants experienced heart attacks or strokes. The study was small and did not measure long-term clinical outcomes.

Those findings justify further investigation. They do not establish the risk from every product, smaller dose or occasional use. The FDA says the observational evidence in the 2023 paper did not establish a causal link between consuming erythritol and the observed effects. Neither dismissing the signal nor declaring universal harm reflects those limits.

Xylitol: a separate ingredient, a separate evidence file

A 2024 European Heart Journal study examined two cardiac-evaluation cohorts totaling 3,306 people and found an association between higher circulating xylitol and subsequent cardiovascular events. The researchers also conducted mechanistic experiments and a short-term ingestion study in 10 healthy volunteers.

The findings raise research questions about platelet responsiveness and exposure. They do not demonstrate that chewing one piece of xylitol gum causes a cardiovascular event. The body can also produce xylitol; long-term dietary effects cannot be read directly from a fasting blood concentration. The authors called for further cardiovascular safety research.

What changed in U.S. oversight

On May 12, 2026, the FDA finalized its food chemical post-market assessment program. Its framework addresses signals about hazards, use and exposure, followed by prioritization, assessment and risk management. Establishing a review process is not the same as completing an assessment of every ingredient.

An FDA proposed rule published August 11, 2026 would require notices for GRAS uses, subject to the proposal's scope and exceptions. As of this article's October 8, 2026 review, it is a proposal, not a final notification requirement. Mandatory notification and FDA premarket food-additive approval are also different concepts.

The practical question is precise: what pathway applies to this use, what evidence supports it and has a reassessment changed the position? An announcement alone should not be described as a ban or a completed safety determination.

Five questions to ask before a headline becomes a verdict

1. Which compound was studied? Do not transfer an aspartame finding to every sweetener or a polyol study to an unrelated ingredient.

2. What kind of evidence is it? Observational studies, cell experiments, animal models and human interventions have different strengths and limits.

3. What was the exposure? Check the amount, duration and population. A large test dose is not automatically comparable to every food serving.

4. What outcome was measured? Platelet activity, body weight and actual cardiovascular events are different endpoints.

5. What remains uncertain? Look for confounding, small samples, endogenous production and missing long-term outcome data. Uncertainty belongs beside the finding, where a reader can see it.

Where IngredientIQ fits

IngredientIQ is developing a food label decoder. The app is not publicly available. The free Smart Label Decoder guide is available today. Planned food label features are not diagnostic services, clinical decision tools or proof that a food is safe for an individual.

The goal is clearer ingredient questions and traceable evidence. No scanner can infer an undisclosed quantity or guarantee a personal health outcome from a label image. A food scanner or AI food scanner should make those limits understandable alongside its output.

Read the free Smart Label Decoder guide or join the early access list.

What to do with this information

Use the ingredient list to identify the sweetener, then read compound-specific evidence. Compare products without assuming that "zero sugar" settles their overall nutritional value. For decisions involving diabetes, PKU or another condition, bring the actual product and your questions to a qualified clinician or dietitian.

The useful question is not whether all sweetness is good or bad. It is what we know about this ingredient, at this exposure, for this person, and where the evidence stops.