Research first
Every exercise page is built from a research doc that covers setup, the movement joint by joint, the muscles involved, mistakes, who should modify it, risks and programming. Every factual sentence in the doc has a source or is labelled as coaching practice. When facts change, the research doc is updated first and the page is rebuilt from it.
Which evidence we use
We use the strongest source available for each claim, in this order:
- Systematic reviews and meta-analyses of relevant populations
- Randomised controlled trials and well-controlled training studies
- Biomechanics and muscle-activity (EMG) laboratory studies: good for how a movement loads tissues, weaker for outcomes like growth or injury
- Position stands and clinical guidelines
- Expert consensus in textbooks and coaching literature
- Practical coaching experience, clearly labelled as such and never presented as research
Higher muscle activity does not prove more growth, so we say "higher measured muscle activity" unless training studies show growth. We note when a study used only trained men, novices, small samples or short durations, and a single small study never supports more than Moderate certainty.
Certainty levels
| Level | Meaning | Typical basis |
|---|---|---|
| High | Unlikely to change with new research | Consistent meta-analyses or many consistent studies; basic anatomy and mechanics |
| Moderate | Probably right; new research could shift details | A few consistent studies, or strong biomechanical reasoning plus supporting data |
| Low | Plausible but uncertain | Single small studies, mechanistic reasoning, expert opinion |
| Insufficient | We don't know | Conflicting or absent evidence, said plainly |
Citations
Every source lives in one registry with its full citation, a link to the record, its verification status and the date it was checked. A source is cited on the site only after its title, authors, journal and year have been checked against an index record and the finding against the abstract or full text. Sources still marked "to verify" are never cited. Quotes from papers are short and rare; we paraphrase and never reproduce figures.
Use of AI tools
AI assistants may help draft, structure, summarise papers and write code. They are never the source of facts.
- Every factual claim is checked against a verified source by a human before a page is published.
- AI drafts are rewritten in our own voice; no unedited AI text is published.
- AI is never used to invent experience, testimonials, reviewer names or credentials.
Pages that have not yet had that human pass are marked as drafts and hidden from search engines.
Review and updates
Health content should be reviewed by a qualified professional before it is published. Until a reviewer is in place, pages show "Last checked by the author" and the date, never a claimed professional review. Each research doc has a next-review date, normally 12 months after the last one. When a recommendation changes materially, the page shows a dated note saying what changed.
Corrections
Reader corrections are welcome: see contact and corrections. Each correction is logged in the research doc's changelog.