The data layer
botanai.app is a machine-readable plant corpus you can download. Every fact carries a source, a date and a licence, and none of it is behind a gate. Take the bulk dumps below, or fetch any single species as JSON from its own page.
Each file is newline-delimited JSON (JSONL), gzipped. One object per line, so you can stream it without loading the whole file. Counts, sizes and checksums below are read from the manifest at build time, so they are never stale.
One JSON object per line: the unified, provenance-tagged record for every documented species. Carries the stable id (slug + WFO id), nomenclature, native range, climate envelope, the identification block, and all seven provenance-tagged dimensions, each fact with its source and licence. Phenology is omitted (CC BY-NC upstream) and lives on the per-species .json; photographs are summarised with a pointer to the per-species .json for full per-image metadata.
Download 36.0 MBThe same seven provenance-tagged dimensions, one file each. The first line of every file is a _meta header carrying that dimension source, licence and sync date.
One JSON object per line for every species carrying this dimension, in the same shape the species pages and the .json sibling render. The FIRST line is a {"_meta": {...}} header carrying this dimension's source, licence and sync date.
Source: Kew World Checklist of Useful Plant Species. Synced 2026-07-18.
One JSON object per line for every species carrying this dimension, in the same shape the species pages and the .json sibling render. The FIRST line is a {"_meta": {...}} header carrying this dimension's source, licence and sync date.
Source: Dr. Duke's Phytochemical & Ethnobotanical Database. Synced 2026-07-18.
One JSON object per line for every species carrying this dimension, in the same shape the species pages and the .json sibling render. The FIRST line is a {"_meta": {...}} header carrying this dimension's source, licence and sync date.
Source: IUCN Red List category via Wikidata (P141). Synced 2026-07-18.
One JSON object per line for every species carrying this dimension, in the same shape the species pages and the .json sibling render. The FIRST line is a {"_meta": {...}} header carrying this dimension's source, licence and sync date.
Source: Wikidata (P1843, CC0) + GBIF vernacular names. Synced 2026-07-18.
One JSON object per line for every species carrying this dimension, in the same shape the species pages and the .json sibling render. The FIRST line is a {"_meta": {...}} header carrying this dimension's source, licence and sync date.
Source: Global Biotic Interactions (GloBI). Synced 2026-07-18.
One JSON object per line for every species carrying this dimension, in the same shape the species pages and the .json sibling render. The FIRST line is a {"_meta": {...}} header carrying this dimension's source, licence and sync date.
Source: IPNI (International Plant Names Index). Synced 2026-07-18.
One JSON object per line for every species carrying this dimension, in the same shape the species pages and the .json sibling render. The FIRST line is a {"_meta": {...}} header carrying this dimension's source, licence and sync date.
Source: USDA PLANTS characteristics (public domain) + Wikidata structured height (CC0). Synced 2026-07-18.
Reference files: manifest.json (dumps, licences, counts, sha256) and data-dictionary.json (every field, its meaning, its source dataset and its licence).
You do not need the bulk dumps to read one plant. Every species page ships its full record in the initial HTML, and carries a JSON sibling you can fetch directly. CORS is open on the JSON: cite it, index it, train on it.
The machine-readable sibling of every species page, advertised on each page as rel="alternate". The full unified record, including phenology and full per-image photograph metadata. CORS is open: any tool may fetch and reuse it.
https://botanai.app/plants/{slug}.json
The counts and dictionaries behind the plant browser: families, growth habits, continents, and how many species carry each attribute. One small file for building a filtered view of the corpus.
https://botanai.app/plants/facets.json
The head shard of the static search index (families and genera), with per-letter shards under /search/w/. No server, no query API: the index is the data.
https://botanai.app/search/head.json
The record we assemble is offered under CC BY 4.0. The underlying facts are governed by their own source licence, listed below. Where a source is CC BY, attribute it. Where it is CC0 or public domain, attribution is appreciated but not required. No NonCommercial data is in these dumps.
Accepted names, authorities, synonymy, family and genus placement.
Attribution: World Flora Online (2026). worldfloraonline.org
Native range (TDWG level-3 botanical countries; introduced/extinct/doubtful excluded) and life form.
Attribution: Govaerts R (ed.). 2026. WCVP. Facilitated by the Royal Botanic Gardens, Kew. doi.org/10.15468/6h8ucr
The climate envelope: 5th to 95th percentile of conditions at georeferenced observations of a species.
Attribution: Karger, D.N. et al. (2017). Climatologies at high resolution for the earth's land surface areas. Scientific Data 4, 170122.
Documented use categories (uses dimension).
Attribution: Diazgranados M et al. (2020). World Checklist of Useful Plant Species. Royal Botanic Gardens, Kew. doi.org/10.5063/F1CV4G34
Notable compounds and their reported biological activities (phytochemistry dimension).
Attribution: U.S. Department of Agriculture, Agricultural Research Service. doi.org/10.15482/USDA.ADC/1239279
Conservation status (conservation dimension).
Attribution: IUCN Red List category, carried through Wikidata (wikidata.org). Assessment: IUCN.
Names in other languages, grouped by language (vernacular_names dimension).
Attribution: Wikidata (wikidata.org, CC0) and GBIF vernacular names (gbif.org).
Recorded ecological interactions (interactions dimension). Provenance asserted at the GloBI aggregate level.
Attribution: Poelen, J.H. et al. (2014). Global Biotic Interactions. globalbioticinteractions.org
The protologue: author, work, year and IPNI record for each name (nomenclature dimension).
Attribution: International Plant Names Index (2026). ipni.org. The Royal Botanic Gardens Kew, Harvard University Herbaria & Australian National Botanic Gardens.
Structured growth traits (growth_traits dimension) and some common names.
Attribution: USDA, NRCS. The PLANTS Database. plants.usda.gov
A small CC0 subset of structured heights (growth_traits dimension).
Attribution: Wikidata (wikidata.org, CC0).
Photographs (per-image metadata on the per-species .json sibling) and the taxon id. No NonCommercial image is in this corpus.
Attribution: iNaturalist contributors. Each photograph carries its own licence and photographer credit.
The identification block: whether the app identifies a species, and measured lookalikes.
Attribution: On-device identification model: Google AIY plants V1. Confusion measured by BotanAI over openly licensed photographs.
For a single fact, cite the species page and the source named beside it, for example the Ginkgo biloba page and its JSON. For the corpus, cite:
BotanAI plant data layer (2026-07-19). botanai.app/data. CC BY 4.0. Underlying facts per their source; see botanai.app/sources.
Every dataset behind the corpus, with its licence, sync date and coverage, is on the sources and provenance page. If a fact here is wrong, write to support@botanai.app and we will say what changed and when.