This page explains exactly how every article on Bio-Hacking.Blog is researched and fact-checked. The short version: every cited claim is checked against a real, reachable source before it is published, and AI does most of the work, as the section on how AI is used explains.
Our source hierarchy
We weight evidence by quality. In order of preference: meta-analyses and systematic reviews, then randomized controlled trials, then cohort and observational studies, then reputable expert reviews. We draw from PubMed, PubMed Central, journal sites, and DOI links, and we tell you when the evidence is preliminary.
How an article is built
AI agents carry out each of these steps, working to rules Lee sets. The next section says exactly where AI is used.
- Plan. We map the question and the specific, falsifiable claims that will each need a citation.
- Research. For each question we open the actual studies and confirm the source genuinely reports the finding. Anything we cannot open and verify is dropped.
- Draft. We write only from the verified sources, with the concrete numbers linked inline.
- Adversarial verification. A separate check re-opens every citation and rules on it: does it support the claim, is it weaker than implied, is it off-topic, or is it dead or fabricated. This step is the whole point, it is what keeps invented studies out.
- Finalize. Fabricated or off-topic citations are removed, weak ones are softened to match what the study actually shows, and the article is published only when the citations hold up.
How AI Is Used Here
Articles on this site are researched and drafted by AI agents, and the process runs automatically. Here is who does what.
- Lee sets the direction. Lee decides what the site covers, and AI agents help fill the queue of topics. Lee also sets the rules every article must follow: the house style, the source standards above, and the plain-language rules.
- AI agents research and draft. Research agents look for studies and must open every source before they use it. Anything they cannot open and confirm is dropped. A writing agent then drafts the article from those confirmed sources only.
- A separate agent checks every citation. A verification agent that did not write the draft opens each cited source again and judges whether it supports the sentence. Dead, fabricated or off-topic citations are removed, and weak ones are softened to match what the study shows.
- A plain-language rewrite. Another pass rewrites the article so it is easier to read. If that rewrite loses even one citation, it is thrown away and the checked version is kept.
- An automated quality gate. A script checks reading level, sentence length, dashes, clickbait wording, the FAQ and the disclaimer. It also blocks statistics most readers cannot picture, such as hazard ratios and p-values. The daily process is set up to rewrite an article that fails and check it again before publishing.
- Then it publishes. New articles go live automatically, usually one a day. No person reads each article before it is published.
The checker’s rulings are recorded for 57 of our articles. As of October 2026, it had judged 1,517 citations in them. Of those, 1,335 held up. Another 180, about 12 in every 100, were weaker than the sentence claimed, so the wording was softened. Only 2 were dead, fabricated or off-topic, and those were removed or replaced. The research step had already dropped any source it could not open and confirm.
What this cannot catch. The checks are real, but they have limits:
- The checker looks at one citation at a time. It can confirm that a study says what a sentence claims, yet miss a misleading overall picture, such as good evidence that was left out.
- It is AI checking AI, so both can share the same blind spot. The checker may see only the summary of a study, not the full paper.
- When the checker rejects a citation, the final edit can swap in a replacement source it suggested. That replacement is not checked a second time.
- Sentences without a citation, such as general background, are not checked against a source.
- The quality gate checks how an article reads, not whether it is true.
- Science moves on, and articles are not checked again automatically after they are published. When an article is edited later, for example to add a newer study, an AI agent makes the edit and it does not go back through the full process above.
- A few of our oldest posts were written before this process existed. Many articles from before late July 2026 were written before the plain-language rewrite and the quality gate were added.
If you find a mistake, please email lee@stutz.science with the article and the claim. We will check it against the source and correct it.
What we will not do
We do not invent studies, statistics, authors, or links. We do not promise cures or guaranteed outcomes. We do not pad word count with claims we cannot back. If a popular idea is not supported by good evidence, we say so.
Corrections
Science moves, and we get things wrong sometimes. If you spot an error, email lee@stutz.science and we will fix it.
Bio-Hacking.Blog is for educational purposes and is not medical advice. Talk to a qualified clinician before changing your health regimen.
