How the data is built
Every plant record in Houseplantdex is built from three sources, each handled with different rules. This page explains them.
Taxonomy — GBIF
The scientific name, authorship, family, genus, and stable ID
(gbif_usage_key) come from
GBIF's species API. GBIF resolves taxonomic
synonyms automatically — when you ask for Sansevieria trifasciata
(the old snake-plant name), it returns Dracaena trifasciata (the
currently accepted one). We always store the accepted name and record
the resolution path.
Toxicity — tiered authority model
Whether a plant is toxic to cats, dogs, or horses is resolved against a ranked set of veterinary toxicology authorities. The first source with a clear verdict for the species (or its accepted synonym) wins; the others are recorded as supporting sources.
- ASPCA — primary authority for cats, dogs, and horses. Species- or genus-level matches only.
- Pet Poison Helpline — used when ASPCA has no entry. Run by veterinary toxicologists; covers many species ASPCA omits (notably the ZZ plant).
- MSD / Merck Veterinary Manual — used as a cross-reference and as the basis for family-level chemistry citations.
- Family-chemistry consensus — used only when the toxic principle is a well-documented family-wide compound (e.g., insoluble calcium oxalate raphides in Araceae, cardiac glycosides in Apocynaceae). Requires at least two veterinary toxicology sources naming the family-level mechanism.
Each toxicity record carries a toxicity_source (which
authority provided the primary verdict) and a
toxicity_match_level (species, genus,
family_consensus, or multi_source). A
supporting_sources array lists every reference consulted,
with the verdict each one gave. All of this is shown on the plant page.
The rules we do not break:
- We join on scientific name, not common name. "Snake plant" can mean a dozen unrelated species; only the binomial is reliable.
- Genus-level inference requires uniform genus toxicity. We only infer from genus when every species in the genus that any authority has assessed agrees.
- Family-consensus calls require a documented chemical mechanism — not just "siblings are toxic." If the family contains both toxic and non-toxic genera with no shared principle, we don't reach.
- "Pet-safe" on cluster pages means ASPCA-verified only. The /pet-safe/ pages list plants confirmed non-toxic by ASPCA — not by inference from any other tier.
Care attributes — generated, then verified
No single open source publishes structured care data for every houseplant. We generate light, water, humidity, temperature, soil, fertilizer, problems, and propagation values, then verify each one against two independent reputable references — preferring the Missouri Botanical Garden Plant Finder, the Royal Horticultural Society, and university extension services (NC State, Iowa State, etc.).
For each field, the verification record notes the value seen at each source and the agreement status. Where sources genuinely disagree, the field is marked disputed rather than picking a winner.
What we won't claim
- That a plant is safe for pets based on anything weaker than ASPCA's non-toxic list. Family-consensus and multi-source reasoning is used to call toxicity, never safety.
- That a plant is toxic just because a sibling species is, absent a documented family-wide chemical mechanism.
- A precise watering frequency (every reference qualifies it with conditions).
- An indoor height for a plant we haven't independently checked.
Updates
The dataset is versioned in a Git repository. Each plant record carries
a provenance block with fetch and verification dates so
anything stale is visible.