Standards

Methodology

Healthspan Post is produced by an AI research system. This page explains how a piece goes from published research to an article, and what is checked along the way.

1. Research

Each day the system searches PubMed, the US National Library of Medicine's index of biomedical literature, and saves the matching papers with their full metadata: authors, journal, year, DOI, study type and abstract. Only papers saved in this step may be cited in that day's writing.

2. Evidence grades

Every research claim is tied to a study, and we say how strong that study is:

• A randomized controlled trial in humans, or a meta-analysis of human trials.
• B small or pilot human study, including uncontrolled or observational studies.
• C animal or cell study. These are always labelled as such, because results in mice or cells often do not carry over to people.

We also report sample size, duration and funding where they matter.

3. Automated checks

Before any human sees a piece, software checks it against our rules. It flags:

• citations to papers that were not in the day's saved research, citations whose details do not match the source, and retracted papers;
• animal or cell findings that are not labelled as such;
• disease claims near a product, testimonials or personal results, and claims that a supplement is FDA-approved;
• a missing affiliate disclosure, affiliate links not marked as sponsored, and altered affiliate links.

A second AI model then reviews each claim against its cited source and marks it supported, overstated or unsupported. Overstated claims are rewritten and unsupported ones are removed. A piece that still fails is held and not published.

4. Human approval

A human operator reads the checks and the claim review and approves or rejects each piece before it goes live. This is an editorial sign-off, not a medical or expert review.

5. Corrections

When we get something wrong, we fix it and record the change on Corrections.