Seven Labs, One Stool Sample, Seven Different Answers
US government researchers split a single stool sample and sent identical material to seven at-home microbiome testing companies. Of 1,208 microbial groups found, three appeared in every result — and one lab called the sample unhealthy while six called it healthy.
Researchers at the US National Institute of Standards and Technology sent identical stool material to seven direct-to-consumer gut microbiome testing companies using 21 kits, publishing in Communications Biology in February 2026. Of 1,208 unique microbial groups identified in total, only three appeared in every company's results, and the variation between providers matched the scale of biological variation between different people. On the same sample, one company reported an unhealthy microbiome while six reported healthy. NIST attributes the spread to methodological variability and insufficient quality control rather than misconduct.
A gut microbiome test costs somewhere between one and three hundred dollars, arrives as a kit, and comes back with a verdict: your diversity score, your ratios, a list of what is flourishing and what is depleted, and often a set of supplement recommendations that follow conveniently from the findings. The obvious question is whether any of it is true. Researchers at the US National Institute of Standards and Technology ran the experiment that answers it, and the result is worse than sceptics expected.
The experimentWhat did NIST actually do?
They took a single standardised human stool sample, split it, and sent identical material to seven direct-to-consumer gut microbiome testing companies — ordering three kits from each, 21 kits in total. Every company was analysing the same stool. Any difference in the results could not come from the sample, because there was only one.
This is the cleanest possible design for the question, and it is the design the industry has never had to face. When two people get different microbiome reports, it is impossible to tell whether that reflects two different guts or two different laboratories. Splitting one sample removes the biology and leaves only the measurement. The findings were published in Communications Biology in February 2026 and written up by NIST in June.
The headline numberHow much did the results differ?
Across the analyses, 1,208 unique microbial groups were identified in total — and only three of them appeared in every company’s results. Three. From one sample. Everything else was found by some laboratories and missed by others.
The comparison NIST draws is the one that should end the conversation: the variation between providers was on the same scale as the biological variation between different people. Put plainly, sending your sample to a different company changes the answer about as much as sending a different person’s sample would.
One stool sample. 1,208 microbial groups found in total. Three of them appeared in every company’s results.
The verdict problemDid they even agree on whether the gut was healthy?
No. On identical material, one company reported the microbiome as unhealthy while the other six reported healthy gut microbiota for the same person. The customer receiving that first report would have had every reason to believe something was wrong, and would have been handed a problem that six other laboratories could not find.
This is where the consumer harm sits, and it is worth being precise about why. A wrong species-level abundance figure is an abstraction most people cannot act on. A verdict — healthy, unhealthy, imbalanced, depleted — is something people absolutely act on, usually by buying whatever the report recommends next. The verdict is the part of the product that is least defensible and most monetised.
The causeWhy do the labs disagree so badly?
Methodology, not fraud. Different laboratories handle samples differently, use different extraction and sequencing methods, compare against different reference databases and apply different thresholds — and there is no shared standard forcing those choices to line up. NIST attributes the spread to methodological variability and insufficient quality control, which is a measurement-science problem rather than an accusation of bad faith.
That distinction matters for how you read the result. These companies are not necessarily doing anything dishonest. They are doing genuinely difficult laboratory work in an area with no agreed reference standard, and then reporting the output with a confidence the underlying measurement does not support. NIST exists precisely to build the standards that would fix this; the study is, in effect, a demonstration of why they are needed.
The practical readShould you buy a gut microbiome test?
For curiosity, knowing the result is not reproducible: that is your call and it is a fair one. For any decision that matters — changing your diet substantially, buying a recommended supplement stack, or concluding that something is wrong with you — the evidence says the result is not stable enough to carry that weight. Reordering from a different company could reasonably return a different verdict.
The specific thing worth resisting is the pipeline from test to purchase. When the same company that scores your gut also sells the supplements that fix the score, the incentive runs in one direction, and the measurement underneath is now documented to be unreliable. That is the combination to be most careful with.
None of this touches diagnostic testing ordered by a clinician for a specific question — a stool test looking for a pathogen, blood in stool, or inflammation markers is a different kind of test with a defined target and established performance. The NIST work is about consumer wellness kits that profile a whole community and grade it. If you have persistent symptoms, that is a reason to see a doctor rather than to buy a kit; as Science News put it in its coverage, the question of whether you can trust the results is currently answered: maybe not. This article is educational and is not medical advice.