Last reviewed 26 August 2026
The compound library on this site carries over six hundred and fifty citations. Before any of them was published, each was resolved against its primary record — the actual PubMed entry, the actual PubChem record, the actual judgment, the actual regulatory document. This page sets out how that works and, more unusually, what the process has caught in our own drafts before publication. A verification process you cannot see catching anything is just a claim; ours has a record, and some of it is embarrassing enough to be believable.
The rule
One sentence governs everything published in the library and the research section: a citation in a draft is a draft, not a fact. Every identifier is resolved against its primary source before publication, and a claim that cannot be verified is softened to what the evidence actually supports, or removed — never left standing because it reads plausibly. The rule exists because modern drafting tools produce text that is confidently formatted and sometimes wrong, and the errors that survive proofreading are precisely the ones sitting in authoritative-looking reference lists.
What gets checked, against what
Journal citations are resolved by PMID against the PubMed record itself — author list, title, journal, year, and the study type PubMed actually assigns, not the one the draft assumed. Chemical identity — molecular formulae, average and monoisotopic masses, CAS and CID numbers — is checked against PubChem. European case law is checked name-against-number in the official reports, including citation form. Regulatory statements are checked against the issuing body’s own document, fetched and read, not summarised from secondary coverage. Plain URLs are resolved; where a site blocks automated access, the underlying identifier is verified through another route and the limitation noted.
What the process has caught — in our own drafts
This is the section no supplier publishes, and the reason to trust the rest. In the library’s first drafting pass the verification stage caught, among other things:
- A fabricated author list on a real paper. A citation carried the correct title, journal, year and pages — with an author list partly spliced in from a different paper by the same group, including a name that does not appear on the cited paper at all. Everything about it looked right; only the primary record showed otherwise.
- A regulatory outcome stated backwards. A draft described a United States compounding decision as the opposite of what happened, attributed to the wrong body in the wrong year. The published entry now states what the record states.
- A self-contradictory anti-doping classification — a compound placed in two WADA categories that are mutually exclusive by definition. The published entry carries the classification the current Prohibited List actually gives.
- Claims contradicted by their own sources. Two entries asserted that a study’s animal model was “not stated” — when the species was in the source’s own title or abstract. The drafts had truncated the titles and then reasoned from the truncation.
- Miscounts, misattributed reporting, an overstated study type, and an unsourced strain name — each small, each corrected, each the kind of detail that makes a reference list quietly unreliable.
Every one of these was caught before publication, and later drafting batches — written under the tightened instructions the first pass produced — came back with zero unverifiable citations across roughly six hundred and twenty checks, though the independent verification pass still corrected nineteen details of year and model against the records. The pass stays, permanently, because the first list above is what happens without it.
What happens when verification fails
A claim that cannot be verified does not ship as written. It is softened to what the evidence supports — “no authorisation identified” rather than “none exists”, because an absence we could not prove is not a proven absence — or it is removed, or, where a detail is genuinely undeterminable, the entry says so in terms. Where a primary text is inaccessible (some pharmacopoeial standards are paywalled), the entry states what it relies on rather than implying a verification that did not happen.
Where you can see it working
Every library entry carries its evidence panel and full reference list — each study linked to its PubMed record, so our characterisation of a paper can be checked against the paper in one click. Research articles carry their sources the same way, and every article now shows the date it was last reviewed. And the standing offer printed across this site applies here with full force: if any statement appears out of step with the record it cites, the record wins — tell us at support@novovita.health and we will correct it.
What this page does not claim
Citation accuracy is a statement about our documentation, not about any compound. A correctly cited study is still a study in its own model and nothing more, and nothing NovoVita supplies is for human or veterinary use — everything is for laboratory research only. The verification process makes the library honest; it does not make anything it describes a medicine.
Sources
- PubMed and PubChem — the primary records journal citations and chemical identity are resolved against.
- EUR-Lex and the official European Court reports — legislation and case law.
- The issuing regulators’ own documents — MHRA, HPRA, FDA, WADA — for every regulatory statement.
- NovoVita, How to verify a UK peptide supplier — the buyer’s-side companion to this page.