Skip to main content

Editorial Policy

CreatineIQ publishes dosing guidance and research summaries about a supplement people take daily. That is health content, and it deserves an explicit account of how it is made. This page covers who writes it, where artificial intelligence is and is not involved, what review happens before anything goes live, and how to tell us we got something wrong.

Who publishes CreatineIQ

Articles are published under the CreatineIQ Research Desk, an organisational byline rather than an individual one. The site is an independent project, not a media company and not a supplement brand, and it is operated by a single owner-editor who reviews everything before it is published.

We publish under an organisational byline rather than a personal one, and we would rather be direct about the trade-off that involves. A named author with clinical credentials is a genuine signal of authority, and we are not offering you one. What we offer instead is a fully documented process — a published research methodology with the actual scoring rubric, an auditable record of which studies were considered and rejected, and a primary source linked behind every claim so you can check it rather than take our word for it.

Nobody here is a physician, dietitian or clinician, and nothing on this site is medical advice. We summarise published research and compute doses from published pharmacokinetics. Decisions about your own health belong with a qualified healthcare provider who knows your history.

How our content is produced

We publish two kinds of article, and they are made differently. Both are labelled on the site: research summaries carry a "New Research" tag and a link to the source study, and guides carry a "Guide" tag.

Research summaries — AI-assisted, human-reviewed

Our weekly summaries of newly published creatine studies are produced by an automated pipeline that uses a large language model at two steps: scoring candidate studies against our published rubric, and drafting the summary of the winning paper. We think you should know that, so we are saying it plainly rather than burying it.

Several constraints are built around those two steps. The model never chooses what is eligible — hard pass/fail gates do that before scoring begins. The model never computes the ranking either; it returns individual criterion scores and the weighted total is calculated in code, so the ranking is reproducible. Study identifiers are never written by the model: the PubMed ID, DOI and journal attached to each article are taken directly from the PubMed record, which is what prevents invented citations. Drafts are then checked automatically against a set of structural and factual rules, and a draft that fails gets one chance to be corrected before the run is abandoned.

Guides — written by hand

Our evergreen guides — dosing, loading, supplement forms, safety, and the rest — are written directly rather than generated. They draw on the same primary literature cited throughout the site, and every substantive claim in them traces to a reference listed at the foot of the article.

Nothing publishes itself

This is the part that matters most. The research pipeline does not write to the live site. It opens a pull request for the editor to read, containing the draft article, the full candidate ranking with per-criterion scores, and the reason every rejected paper was dropped. A human reads it and merges it, or does not. There is no automatic merge and no scheduled auto-publish.

Studies flagged as potentially groundbreaking are held to a stricter version of the same rule: they are generated as unpublished drafts with no public URL at all, and only become visible after the editorial rationale has been proofed by a person.

Sourcing standards

  • Primary literature, linked. Research summaries link to the PubMed record and DOI of the study they describe. Claims in guides are traceable to a listed reference.
  • Identifiers are never invented. PubMed IDs, DOIs and journal names come from the PubMed record itself, not from a model's recollection. Volume, issue and page numbers are omitted rather than guessed.
  • Conflicts of interest are reported. Where a study discloses industry funding or author relationships with supplement manufacturers, the article says so — including when the disclosure sits with a prominent researcher, and including when it points at a result we would otherwise be inclined to like.
  • We criticise the studies we cover. Small samples, short durations, biomarker-only endpoints and conclusions that outrun the data are called out in the article itself. Coverage is not endorsement, and a study appearing here is not us telling you it is good.
  • Uncertainty is stated. Where evidence is preliminary — a narrative review, a pilot trial, a mechanism without an outcome — the article says so rather than presenting a hypothesis as a finding.

Corrections

We would rather be corrected than be wrong quietly. If you find a factual error, a misread study, a broken citation or a claim that overstates its evidence, email editorial@creatineiq.com or use the contact form and select "Correction / Factual Error", which routes to the same inbox. Point us at the article and, where you can, the source that shows the error — that makes it far faster to act on.

For anything that is not a correction — feature ideas, bugs, general questions — write to contact@creatineiq.com.

How we handle them:

  • Substantive errors are fixed at the source. If a finding, a number or a dose is wrong, we correct the article itself rather than appending a note and leaving the error in place.
  • Corrections that change the meaning are disclosed. Where a fix materially alters what an article claimed, the change is noted in the article so a returning reader is not silently shown something different.
  • Reviewed articles are dated. When an article has been checked against its sources, it carries a "Last reviewed" date in its header. Articles without that line have not yet had a recorded review — we would rather show you nothing than a review date we cannot stand behind.
  • Process failures change the process. Where an error came from our pipeline rather than from a one-off mistake, we change the pipeline's rules so the same class of error cannot recur.

Independence and funding

CreatineIQ is funded by display advertising and by voluntary reader support. Advertising is served by a third-party network; we do not choose the individual adverts shown, and advertisers have no input into what we publish or how a study is scored.

We do not publish sponsored articles, and we take no payment from supplement manufacturers for coverage. No brand has any influence over the rubric, the rankings, or the recommendations in our guides — which is why those guides recommend the cheapest form of creatine on the shelf. If we ever introduce affiliate or sponsored placements, they will be visibly labelled as such and kept out of the editorial judgement above.

Calculations run entirely in your browser. Your body composition and health inputs are never transmitted to us or stored on a server — see our Privacy Policy.

Not medical advice

Everything on CreatineIQ is for education and research reference. It is not a substitute for professional medical advice, diagnosis or treatment, and our brain health protocols — including anything touching concussion or traumatic brain injury — are not emergency care plans. Consult a qualified healthcare provider before starting any supplement, particularly if you have kidney or liver disease, are pregnant or breastfeeding, are under 18, or take medication affecting kidney function.