Roman Sydorenko
Independent analyst and operator of this site. His work is record-checking — resolving strain designators to published records, chasing certifications to the issuer’s own register, and testing whether a published number survives being looked up. Not a clinician; every clinical claim here is sourced to a named trial, guideline or regulator record you can open yourself.
What that background is good for
My work is search and data analysis: finding records, reading them against each other, and checking whether a published number survives being looked up. That is the entire skill this site rests on, and it is genuinely useful here, because most of what is wrong with this category is not clinical — it is a strain designator that resolves to nothing, a CFU count with no temporal basis, a certification badge that is not in the certifier’s register, a percentage that turns out to come from a survey of paid customers, and a trial the marketing evokes that was run on a different strain at a different dose. None of those is a clinical error. They are record errors, and they are visible to anyone willing to open the record.
A relevant interest, disclosed
I run an independent SEO consultancy. That is the unrelated work that pays for this site, and it is a real interest a reader should know about: somebody whose living comes from search visibility has an obvious incentive to make a site look authoritative, and this site is designed to rank. I am not going to argue my way out of that with a character reference. The defence on offer is structural — every number here carries the record it came from, the method is published, and the audit cohort was frozen and dated so that no percentage on this site can be quietly re-cut later to suit anyone.
Where this background stops
I am not a clinician, and I do not give medical advice — not here, and not anywhere on this site. Record-checking is not clinical judgement. Where a claim needs clinical judgement, this site quotes a named guideline body verbatim so the judgement on the page is theirs and attributable, and where a reader needs an examination, no page here is a substitute for one.
No degree, licence number, ORCID or institutional affiliation appears on this page, because none is relevant to what this site does. The usual fix in this category is to buy one — a “medically reviewed by” line attached to somebody who never read the page. This site does not have one, and its review status says so rather than implying otherwise.
How the work is checkable
The claim this site makes is not “trust me”. It is that you should not have to. Four things carry that, and each one is a thing you can test rather than a thing you are asked to believe.
Every claim traces to a primary source with a resolvable identifier
Not a link to another article about a study — the identifier itself: a PubMed ID, a trial registry number, a DOI, an NIH label-database record, a regulator’s own docket or register entry. Trials are read for the numbers actually reported, including the ones the abstract does not lead with: missed primary endpoints, null secondaries, sample sizes after exclusions, and who paid. If you want to check a claim, open its identifier and read the Results section, not the Conclusions; on this material the two disagree more often than you would expect.
The method is public before the findings are
The methodology page publishes the three denominators, the teardown procedure step by step, the grading definitions, the correction policy and the re-verification cadence — including the databases that were not searched, which is the scope every absence statement here is limited to. A method published after the conclusion is a rationalisation.
Nothing here is published on my say-so
Every draft on this site is read adversarially before it is published, by a pass whose only job is to break it. Not proofreading: it re-runs the queries, re-opens the papers, and re-checks the registers against what the draft says they contain. Author names are read off the record rather than composed from context, conflict-of-interest statements are transcribed rather than characterised, quotations are checked character by character, and a count that no re-run of its own stated query reproduces does not publish. The methodology page sets out what that pass checks.
That is a description of the method, not a score. What is published is what survived, and where something published here turns out to be wrong it goes in the corrections log with the original wording quoted.
The corrections go against me too
Corrections here are dated, permanent, and quote the original error in full. Nothing on this site is silently edited — if a number here is wrong, the wrong version stays visible next to the right one, with my name on both.
Check something, and tell me what is wrong
This is the part that makes the rest of it worth anything. Open an identifier on any page and read the source yourself. Re-run an absence statement. Look up a certification in the certifier’s register. If a number does not survive that, it is an error and I want it. Corrections are welcome from anyone, including from a brand whose product is audited here, and they are published on the same terms.
Send corrections to roman@seobro.com. The corrections policy sets out exactly how one is published.
Clinical review of my work
No clinician has reviewed this site. If one ever does, their name, their credential as it appears on the issuing register, and the date of the review will be published on the pages they reviewed. What a reviewer would and would not be asked to check is published in advance, on the disclosures page and on the methodology page.