0About·Methodology

Methodology

Where every line of content on PlantSage comes from, who reviewed it, and how confident we are about it.

Data sources

The catalogue is built from four upstream sources, merged with explicit provenance tracked per field:

Permapeople — permapeople.org. Open botanical database. Source of baseline taxonomy, common names, edible flags, basic cultivation.
Trefle — trefle.io. Global plant API based on TDWG taxonomic data. Source of distributions, multilingual common names, growth specifications (light, water, soil, hardiness, pH, temperature ranges, bloom months).
Scientific references — RCTs, meta-analyses, pharmacopoeia monographs (EMA, WHO), traditional-medicine textbooks. Each medicinal claim links to the specific reference behind it.
Traditional + folk sources — old herbals (Culpeper, Gerard), regional ethnographic literature, family records contributed by users. Marked clearly as traditional, never inflated to clinical evidence.

Editorial review

Every plant page passes through a workflow with explicit states:

Status	Meaning
draft	Imported from upstream sources, not yet reviewed. Not indexed by search engines.
in_review	Editor is actively checking. Public but flagged as in-review on the page.
reviewed	Editor has verified the data against cited sources.
published	Public, indexed, and considered stable. Updates re-trigger review.

III

Confidence ratings

Every medicinal claim carries a confidence rating. We don't inflate.

highSystematic reviews or multiple RCTs

Cochrane reviews, meta-analyses of ≥3 randomised controlled trials, large-N (>200) RCTs.

moderateSingle RCT or strong observational evidence

One quality RCT, large cohort study, or well-documented mechanism with consistent clinical signal.

lowTraditional or in-vitro only

Long-standing traditional use, animal studies, or in-vitro pharmacology. No human clinical evidence.

AI assistance

Some descriptive text on PlantSage is generated by Anthropic Claude Haiku 4.5 from structured botanical data we already have (Latin name, family, energetics, distributions, etc.).

What we use AI for

Plant description paragraphs (when no manual editorial copy exists)
Cultivation summary text (from structured energetics + growing data)
FAQ entries on landing pages
Meta descriptions for SEO

What we never use AI for

Medical claims (always from cited sources)
Dosage recommendations (always editorial)
Safety / contraindications (always editorial + cited)
Inventing facts not in the underlying data

AI-generated content is tracked in our database with model name, generation date, and prompt version. Editors can override any AI-generated field with manual copy.

Disclaimers non-negotiable

Not medical advice

PlantSage is an educational resource. Information about medicinal plants does not constitute medical advice and is not a substitute for consultation with a qualified healthcare professional. Herbs can interact with prescription medication, cause allergic reactions, or be contraindicated in pregnancy or specific health conditions. Always consult a doctor or licensed herbalist before using any plant medicinally, especially if you are pregnant, nursing, taking medication, or have a chronic condition.

Corrections & contact

Spotted an error? Please tell us: corrections@plantsage.eu

Want to contribute? See Submit a remedy or correction →. We credit every contributor.

❦❦❦❦

Endpaper · archive updates

Get archive updates

Occasional email when substantial new content lands — plant profiles, seasonal growing notes, folk-remedy additions. Free, no tracking, unsubscribe anytime.