Kefir is almost always introduced as "a drinkable yoghurt, more fermented". The international standard governing it says something else, in one line: kefir must contain live yeasts, and yoghurt is held to no such requirement.
Everything that follows comes from that regulatory detail: the fizz, the taste, the alcohol, and part of what the product does or fails to do.
What legally separates kefir from yoghurt
The Codex standard on fermented milks defines kefir by its starter, prepared from kefir grains combining Lactobacillus kefiri with species of the genera Leuconostoc, Lactococcus and Acetobacter, in what it calls a strong specific relationship. It specifies that the grains also carry yeasts, lactose-fermenting or not.
The composition table is blunter than the definition. For fermented milk, yoghurt and kefir alike, the standard requires at least 10⁷ starter micro-organisms per gram. It then adds a line reserved for kefir and kumys: at least 10⁴ yeasts per gram. That line is blank for yoghurt.
The consequence nobody mentions: there is alcohol in it
Yeasts fermenting sugar produce ethanol and carbon dioxide. That is basic biochemistry, and it is measurable. A team fermented whole milk for 48 hours with milk kefir grains: 5.17 g of ethanol per litre, with pH falling from 6.70 to 4.99. With water kefir grains on the same milk, less than 0.3 g/L.
In practical terms, 5.17 g/L is about 0.65% by volume. A 250 mL glass carries around 1.3 g of ethanol, roughly eight times less than a half-pint of beer. Negligible for an adult having one glass. Not zero, and the distinction matters for a child, a pregnancy, chosen abstinence, or a religious prohibition.
Lactose, the only ground where kefir has a dedicated trial
Fifteen adults who digest lactose poorly received, fasting, portions each delivering 20 g of lactose: milk, plain kefir, flavoured kefir, plain yoghurt and flavoured yoghurt. The measure is breath hydrogen, a direct marker of undigested lactose fermenting in the colon.
| What was drunk | Breath hydrogen | Gap with milk |
|---|---|---|
| Milk | 224 ppm·h | reference |
| Yoghurt, plain or flavoured | 76 ppm·h | −66% |
| Plain kefir | 87 ppm·h | −61% |
| Flavoured kefir | 156 ppm·h | −30% |
The kefirs and yoghurts all reduced perceived flatulence severity by 54% to 71%, with abdominal pain and diarrhoea negligible throughout. The useful detail is the gap between plain and flavoured: flavoured kefir loses half the effect. The trial did not set out to explain why, so neither will I.
On the rest, the meta-analyses tell different stories
A meta-analysis of 6 trials and 314 participants finds fasting insulin down by 3.69 µIU/mL and HOMA-IR down by 2.56. HOMA-IR is an index computed from fasting glucose and insulin that estimates insulin resistance. Nothing, on the other hand, for total cholesterol, triglycerides, HDL, LDL, fasting glucose, glycated haemoglobin or weight.
A second, broader one takes in 24 studies and also finds improved insulin resistance, with HOMA-IR down by 1.71, plus a fasting glucose reduction of 8.46 mg/dL that the first did not see. It finds nothing on body measures, the lipid panel or inflammatory markers.
The two agree on one point and part ways on another. The robust signal is insulin sensitivity. Fasting glucose depends on which studies are included, the usual sign of an effect too small or too unstable to settle. And a third meta-analysis, across 7 trials and 385 participants, finds nothing: 1.76 mmHg off systolic pressure, 1.19 off diastolic, 0.17 mg/L off CRP, none of it significant.
The microbiome, the most sold promise and the least kept
It is the headline argument on every carton, and it is the weakest ground. A systematic review gathered 8 human intervention studies and concludes there are "modest and heterogenous" changes in microbiome composition, with minor shifts at the level of major bacterial groups and a rise in the genus Bifidobacterium in 3 studies out of 4.
The most interesting point sits elsewhere. Species delivered by kefir are indeed recovered in stool, but the authors speak of transient persistence, not establishment. In other words, as long as you keep drinking it, they are there. They add that direct functional evidence in humans, meaning a health benefit tied to that change, remains limited.
What is actually in the bottle
Five commercial kefirs were sequenced and counted. Among those stating a colony-forming-unit count per gram, 66% overstated microbial density by at least tenfold, only one reached the 10⁹ CFU/g mark, and species that were clearly present, such as Streptococcus salivarius or Lactobacillus paracasei, did not appear on the label. It is the same gap between the wording and the content as on a bottle of olive oil.
The easy reading would be to blame industry and send everyone back to homemade grains. It does not hold. Freeze-dried starters made of five traditional bacteria and four traditional yeasts reproduce a kefir identical to the fresh-culture version: same final pH, same bacterial and fungal counts, same composition on sequencing, and no difference in lactic acid, acetic acid, ethanol, glucose or galactose. The process is not the problem. The declaration is.
Milk kefir and water kefir are not the same product
The two share a name and a principle, living grains, and not much else. The first systematic comparison of the two grain types finds significant differences in microbial content and chemical properties: water kefir grains host greater microbial variety, milk kefir grains are nutritionally denser.
The practical consequence: nothing measured on lactose, insulin or the microbiome in the trials cited here transfers to water kefir. Those trials used milk kefir. Water kefir is an interesting drink, lactose-free and compatible with a vegan diet, but its clinical file is close to empty.
What this article does not say
- That kefir heals the gut. The changes observed in the human microbiome are called modest and heterogeneous by the authors themselves.
- That these trials are large. Six trials and 314 participants, seven and 385: these are small sets, and effects isolated in small sets tend to shrink afterwards.
- That the alcohol in kefir is a health problem. At one glass a day it is negligible. The point is not the dose, it is knowing it is not zero.
- That everyone can drink it. A live product containing yeasts is worth discussing with a doctor in case of immunosuppression.
« Kefir is not a sourer yoghurt. It is a yoghurt that also has yeasts, and everything else follows. »
What Helix does
Kefir is a clean case of a food whose promise and whose file do not overlap. The microbiome is the selling point, insulin sensitivity is the most solid result, lactose is the only area with a dedicated trial, and the alcohol appears nowhere. The engine separates those layers: what was measured, on which outcome, at what sample size, and what that is worth next to the rest of your stack.