Our sources and method

The first rule

Before a piece of data enters the site, its source must be nameable. When no solid source can be found, the data stays out, or it comes in stating its limit.

A source is cited by its issuer and its title, the way a book is cited. It does not recount how it was read, and it is never reduced to a category: we name the work, the page or the article, or we drop the claim.

The order of trust

References. Pharmacopoeias, health agencies and public bodies: the European Medicines Agency, EFSA, ANSES, INRAE, FAO, Tela Botanica, botanical conservatories, national weather services, forest inventories. This is the level required for botany, food safety and regulation.

Published studies. Author, year, journal, and a DOI checked against Crossref before being cited. This is the level required as soon as a sentence puts forward a mechanism, an effect or a figure.

Books, ancient texts and ethnography. For permaculture philosophy, ethnobotany and traditional recipes. An ancient text is cited in an identifiable edition, never from memory.

Recognised practice. Learned societies and technical bodies: the Royal Horticultural Society, ITAB, chambers of agriculture, regional gardening academies, Plants For A Future, naturalist associations. This is the level of growing advice, cross-checked between sources.

Documented tradition. A use that is attested and situated, always labelled as such, never presented as established fact.

General gardening websites. They give practical advice or relay information. They never serve as a health warrant, and they are cross-checked against a higher-level source whenever one exists.

What the site refuses to write

No therapeutic promise. A plant is never presented as able to make an illness go away. Medicinal uses are given for what they are, a traditional use or an official monograph that frames them, with their cautions. Nothing here replaces the advice of a doctor, a pharmacist or a botanist.

No precise figure without a source. No milligrams, yields or distances put forward because they sound right. In doubt, a range, or nothing.

No study cited from memory. A reference whose existence has not been checked is not written. Better to name a tradition than to invent an author and a year.

No invented plant pairing. Companion planting that belongs to garden lore without scientific backing says so explicitly.

Limits, written rather than hidden

A calendar is a landmark, not a date. It shifts with the region, the altitude, the aspect and the way you grow. Each plant page shows, under its season band, the sources for the climate zone you chose, and says when its dates are derived from a rule rather than recorded for that zone.

When two sources disagree, the page says so instead of silently picking one. When only one source documents a plant, that is written too.

An absence is stated as an absence. A page with no known companion says so, rather than leaving a gap that would read as an oversight.

How the rule holds

Automatic checks re-read the data on every change, before anything goes online. An entity with no source, a source that recounts how it was checked instead of citing, a health warrant taken from a gardening website, an announced figure that no longer matches the list it announces, two pages giving contrary advice: each of these makes the check fail.

These guards were born from real mistakes, found by re-reading the site. A book with an inaccurate title, a repealed decree cited for thresholds, a plant whose page stated the opposite of the reference review: every correction was followed by a rule that stops the class of error from coming back.

Found a mistake?

Tell us. The contact address is on the legal notice page. A correction is worth more than a comfortable claim, and corrections received are tracked in the project's public history.