A bottle of kombucha a day for 12 weeks of running: the antioxidant score it was meant to lift didn't move, and the four results that did faded on a stricter test

Kombucha is sweetened tea left to ferment with a mat of bacteria and yeast, and it comes out sour, lightly fizzy and full of the plant compounds that tea is known for. That is the basis of its reputation as an antioxidant drink. Researchers at the Federal University of Viçosa in Brazil, who brew their own for research, wanted to see whether that reputation holds up where it should matter most: in people whose bodies are under fresh physical strain. Their trial appeared online in the Journal of Food Science on 26 September 2026, and they describe it as the first randomised trial of kombucha in runners.

They started with 1,120 people who filled in an online form and ended up randomising 90: healthy adults aged 18 to 45, about 25 on average, at a normal weight, not very active, and not in the habit of drinking kombucha, green tea or other fermented foods. Thirty got a 200 ml bottle of green tea kombucha every day, thirty got the same kombucha with a grape-skin extract stirred in (the team hoped the extra polyphenols would add to the effect), and thirty were the control group. Everyone ran on an athletics track three times a week for 12 weeks under supervision, with the distance creeping up each week. Blood was taken at the start and the end, both after an overnight fast and within ten minutes of running to exhaustion.

The measure the trial was built around is called FRAP, a lab test of how much antioxidant power a sample of blood plasma has. After 12 weeks it was no different in the kombucha groups than in the controls, fasting or after exercise. A single bottle drunk an hour and a half before the first all-out run did nothing measurable either: white blood cells surged and settled the same way in all three groups. The authors say this directly in the abstract, which begins its results with the words 'not associated with significant changes in the primary outcome'.

Four secondary results did cross the usual line. In the plain kombucha group, neutrophils went from about 52% to 59% of white blood cells and lymphocytes fell from 38% to about 31%, shifts the other two groups didn't show (P = 0.026 and 0.023). An antioxidant enzyme, catalase, held up slightly better than in the controls. And a marker of fat damage measured after exercise, MDA, fell by 54% in that group. Each of those comes with a catch the paper spells out. The actual numbers of neutrophils and lymphocytes didn't differ between groups, only their shares. The catalase difference was not backed by the test that asks whether groups changed differently over time (P = 0.157). The MDA result passed that test but no pair of groups differed when compared head to head. And when the team corrected for the number of things they had measured, none of the four stayed significant. They call them exploratory.

The grape-skin version is its own small lesson. It had more phenolic compounds than the plain kombucha, 0.94 against 0.72 mg per ml, and it produced none of the four signals. More of the ingredient that is supposed to do the work, and less to show for it. The authors suggest the plain drink's higher acidity may matter for absorption, but they did not measure how much of anything reached the blood, and say so.

The rest is our reading. Nobody was given a dummy drink, because, as the authors put it, kombucha's sourness is impossible to disguise; everyone knew what they were getting, though blood counts are hard to sway by expectation. Twenty-seven of the 90 dropped out, leaving exactly the 21 per group the sums required, and the losses were lopsided: the grape group went from 18 women to 10. The trial was also sized to catch a large effect, borrowed from a grape juice study in athletes, so a modest benefit could have gone unseen. That is a fair reason not to read this as proof that kombucha does nothing. But the clearest finding in the paper has nothing to do with the drink. Twelve weeks of running changed nearly every marker in every group, and not in the tidy direction supplement adverts imply: hydrogen peroxide went up and catalase went down across the board. Training was the big lever. Each bottle, for what it's worth, carried about 9 g of sugar. The work was paid for by Brazilian public research agencies and the authors declare no conflicts of interest.