Long-form explainers built from peer-reviewed research, UN datasets and government dietary guidance — plus practical how-tos you can use the same day. Every number is attributed, every claim is checkable.
The 7-day starter plan translates everything above into a shopping list and simple recipes.
About this library
Each guide summarises a body of research on one question, states the strength of the evidence honestly, and links to the primary sources. It is written to be checkable rather than persuasive.
Guides prioritise systematic reviews, meta-analyses and large prospective cohorts. Where only observational data exist — which is often the case in nutrition, since randomising decades of diet is impossible — that limitation is stated in the text rather than buried. Mechanistic and animal-model studies are used to explain plausibility, never as standalone proof of a human outcome.
Claims fall into three tiers. Strong: consistent findings across multiple large cohorts and meta-analyses, with plausible mechanism — for example that plant-based dietary patterns lower LDL cholesterol. Moderate: consistent direction with residual confounding or smaller samples, such as effects on type 2 diabetes incidence. Emerging: promising but immature, such as much of the gut-microbiome and mood literature. Presenting all three at the same confidence is the most common failure in popular nutrition writing.
People who eat plant-based diets in Western cohorts also smoke less, drink less and exercise more, which inflates apparent benefit. Good studies adjust for these, but adjustment is imperfect. The same problem also works in reverse: 'healthy user' effects can mask real benefits in populations where plant-based eating tracks with poverty rather than affluence, which is why cohorts outside high-income countries matter.
Every claim carries its source. If a guide misreads a paper, overstates a finding or omits a significant contrary result, that is a correctable error and worth reporting. The measure of a library like this is not that it never gets things wrong but that its errors are traceable and fixed.
| Tier | Criteria | Example claim |
|---|---|---|
| Strong | Meta-analyses + consistent cohorts + mechanism | Plant-based diets lower LDL cholesterol |
| Strong | Large-scale LCA consensus | Beef has the highest footprint per kg of protein |
| Moderate | Consistent cohorts, residual confounding | Lower type 2 diabetes incidence |
| Moderate | Good data, contested attribution | Livestock's share of global emissions (~12–20%) |
| Emerging | Small or short-term studies | Diet effects on depression and anxiety |
| Contested | Genuine scientific disagreement | Soil carbon offsets in managed grazing |
A moderate-tier finding stated confidently in conversation is how good evidence gets a bad name.
Read the abstract and the limitations section. Ten minutes there beats an hour of secondary commentary.
A statistically significant 3% relative risk change is real and usually unimportant. Magnitude decides relevance.
The uncertainty sections are the most informative part of any honest summary.
They are compiled from published literature and public datasets, with sources listed on each guide so any claim can be traced independently of who wrote it.
No. It is a summary of research. Decisions about pregnancy, chronic illness, medication or feeding infants belong with a qualified clinician.
Because omitting them would misrepresent the state of the evidence. Grazing and soil carbon, for example, is genuinely unresolved and is presented that way.
They are reviewed periodically against new meta-analyses and updated datasets; each guide carries the sources it relies on so you can check whether something newer has appeared.
Yes, under the site's Creative Commons licence, with attribution and a link. Sources should be cited to their original publishers rather than to this site.