Probiotics sold to men, label by label

This is the label audit, and this page is its catalogue. It answers one question, 166 times: what does this product’s label say, and what of it survives being checked? One row per product, one column per question, and every cell traceable to a record you can open. What those labels add up to across the audit — the counts, the percentages, the regressions — is the other question, and it has its own page. This one ranks nothing and recommends nothing.

Which products are here is a rule you can re-run, not a shortlist somebody assembled: the universe rule admitted 167 SKUs on 31 July 2026 and 166 of them have a page in this build. A product missing from this table is a product the rule did not admit or that has not been torn down yet — never a product that is not sold.

Roman Sydorenko, Independent analyst · Last verified 30 July 2026 · Disclosures

The table #

If you would rather have this as a file than as a table: the founding cohort is published as CSV and JSON under CC BY 4.0, with a citation line, a data dictionary and the limits of the release stated on the page.

What those 27 labels add up to — strain designators, CFU and potency basis, seven documented label regressions, the pricing finding, the dose gaps and the certification registers — is written up in full at what the labels actually say. This page is the records; that page is what they found.

How to read this table #

Teardown

Whether the underlying record is finished. An incomplete teardown publishes anyway, marked, because hiding it would make this audit look more complete than it is.

Cohort

Which denominator the row belongs to. D₂₇ founding is the frozen, dated cohort of 27 teardowns and 294 organism slots that every percentage on this site divides by. Addendum is a teardown completed after that cohort was frozen; it gets a full page and sits outside those denominators, so the published percentages cannot move under a citation. Do not compute a ratio across both.

Potency basis

What the CFU count is measured against. Only a through-expiry figure is falsifiable by somebody holding the bottle; a manufacture-only figure makes no commitment about what is alive when you swallow it.

Slots with a human RCT

How many of the product’s organism slots carry a strain with a human randomised controlled trial under the exact designator printed on the label. Across the frozen founding cohort, 0 of 294 slots carry an organism with such a trial for any male-specific outcome. That is a D₂₇ figure; this column is per product and every row in this table has one.

Certification at the issuer

Checked at the certifier’s own public register, never at the badge on the bottle. “None checked” means we have not queried a register — it does not mean the product holds nothing, and the two are different findings.

Findings

How many named, cleared findings this teardown carries, each with an identifier on the product’s own page. A zero is not a clean bill of health; it may mean the work is not done.

Published gaps

How many things this teardown wanted and does not have — a register that would not answer, a panel not fully transcribed, a price with no dated source. This column grades us, not the products, and it is sortable for that reason.

What is missing from this table #

Across the 166 records above there are currently 1192 published gaps — fields this audit wanted and does not have. They are itemised on each product’s own page under “what we could not verify”, with the reason for each. None of them is deleted, rounded away, or filled with an estimate.

Coverage and denominators are two different things, and this table is the place they are most easily confused. Coverage is 166 pages against a universe of 167 rows, and it grows. The denominators are frozen and dated: 27 products taken apart at label level and 294 organism slots across those labels. Anything torn down after that cohort was frozen publishes as a dated addendum outside those denominators, so the published percentages cannot silently move under a citation — which also means every percentage on this site describes 27 of these 166 products, not all of them. The reasoning is on the methodology page.