Do gut microbiome tests work? One sample, seven companies, three matches

Not reliably, no. In 2026 researchers at the US National Institute of Standards and Technology sent one identical, homogenised stool sample to seven direct-to-consumer gut microbiome companies, three times each. Of 1,208 microbial groups identified across the 21 results, only three were detected in every single sample. The differences between companies were as large as the differences between different people.

What did the 2026 test actually do?

The study, led by Stephanie Servetas with Diane Hoffmann, Jacques Ravel and Scott Jackson, was published in Communications Biology in 2026.

The design is the important bit, and it is elegant. The team used a standardised human faecal reference material developed at NIST, which means every aliquot was chemically and biologically identical. Any difference in the results could therefore only come from the testing, not from the sample.

They sent that material to seven companies selling gut microbiome tests to consumers, in triplicate, for 21 tests in total.

How different were the answers?

Across all the analyses, 1,208 unique microbial taxa were reported. Only three genera turned up in every single sample. After excluding one outlier replicate, 17 genera were shared across every dataset.

Reported sequencing depth ranged from fewer than 20,000 reads to more than 10 million. The number of species identified ranged from fewer than 100 to more than 1,700, depending on which company you asked.

Most damning for anyone trying to use these results: the variation between providers was on the same scale as the biological variation between different donors. Put plainly, sending one person's sample to two companies produced results about as different as sending two different people's samples to one company.

Some companies also disagreed with themselves. Repeat tests of the identical material from the same provider did not always match.

Why do the results differ so much?

Because every step of the process is a choice, and the companies make different ones.

DNA has to be extracted from the sample, and different extraction chemistries break open different bacterial cell walls with different efficiency. The DNA then has to be sequenced, and the companies vary in method and in how deeply they sequence. The sequence data then has to be matched against a reference database and assigned to named organisms, which involves a pipeline, a database and a set of thresholds for how confident you have to be before you report a name.

Change any of those and the species list changes. The NIST authors attribute the differences to exactly this: methodological variability and a lack of sufficient quality control.

They also note a problem the consumer cannot work around. Many companies give limited information about their analytical methods, which makes comparing results across platforms difficult even for specialists.

Does this mean the gut microbiome does not matter?

No, and the distinction matters.

The science linking gut bacteria to health is a serious and active field. There are genuine clinical uses of stool testing, and faecal microbiota transplant is an established treatment for recurrent Clostridioides difficile infection. Nothing in the NIST study touches any of that.

What the study tests is narrower and more immediate: whether the commercial product you can buy today measures what it says it measures, consistently. It answers that question, and the answer is not yet.

The authors make the logical order explicit. Analytical performance is a prerequisite for making sound clinical recommendations. A test that cannot reproduce its own result cannot support a dietary plan built on top of it.

What do gastroenterologists say about these tests?

In February 2025 an international expert panel published a consensus statement on microbiome testing in The Lancet Gastroenterology and Hepatology, led by Serena Porcari and Benjamin Mullish with a large multidisciplinary group.

Their starting observation is blunt. There is growing interest in using the gut microbiome as a diagnostic tool, but the evidence supporting its clinical usefulness is scarce, and an increasing number of commercial providers sell direct-to-consumer tests without any consensus on their regulation or any proven value in clinical practice.

The panel's stated aim was to establish a framework to regulate the provision of microbiome testing and to minimise the use of inappropriate tests, which tells you how they view the current market.

Why are these tests not regulated like medical tests?

Because of where they sit. These products straddle the line between medical devices, which are strictly regulated, and general health and wellness products, which are barely regulated at all. The NIST authors point out that this distinction may not be readily apparent to consumers.

A test marketed as giving you insight into your wellbeing rather than diagnosing a disease avoids the evidence requirements that a diagnostic test would face. The report you get back can carry species percentages, diversity scores and food lists without anyone having to show that the numbers are reproducible, let alone that acting on them improves anything.

What about the diet advice the report gives me?

Treat it as the weakest part of the product, because it is built on the part that does not reproduce.

The chain of reasoning in a typical report runs: we measured your bacteria, your bacteria differ from an ideal profile, these foods will shift them, shifting them will improve your health. The NIST study breaks the first link. If the measurement varies between providers as much as it varies between people, the personalisation downstream of it is not personal to you.

That does not make the advice harmful. Most of it is some version of eat more plants, more fibre and more fermented food, which is reasonable general advice that you do not need a test to receive. It does mean you are paying for the test and getting the advice. We have written about the same pattern in at-home biomarker panels and in biological age tests.

Is there anything worth doing with a gut symptom instead?

Yes, and it is less exciting.

Persistent symptoms deserve a clinician rather than a kit, because the things that need finding are findable. Blood in the stool, unexplained weight loss, a change in bowel habit that lasts, anaemia, or a family history of bowel cancer all have specific tests behind them: coeliac serology, faecal calprotectin, faecal immunochemical testing, and colonoscopy where it is warranted. Those tests are validated, which is exactly what the microbiome kits are not.

For general gut health, the interventions with reasonable evidence are unglamorous: more dietary fibre, a wider range of plant foods, enough sleep, less alcohol, and not taking antibiotics you do not need. Our round-up of gut health trends goes through what has evidence and what does not.

Should I buy a gut microbiome test?

If you want one out of curiosity, and you treat the output as entertainment rather than information, the main cost is money.

If you intend to make decisions with the result, change your diet because of it, or spend money on supplements the report recommends, the 2026 NIST data is a good reason not to. You would be acting on a number that another company, or the same company on a different day, might not reproduce.

The thing worth watching is standards. The NIST authors' conclusion is that the field needs reference materials and reporting requirements so analytical validity can be demonstrated. When companies start publishing their performance against a standard like the one used in this study, the product becomes assessable. Until then, the honest summary is that nobody, including the companies, can tell you how accurate your result was.

This article is general information and not medical advice. Speak to a qualified clinician about persistent gut symptoms.