Every claim on this site traces to a dataset named below. Where we do not have something, the page that would have carried it says so, rather than filling the gap with prose that sounds right.
Accepted names, authorities, synonymy, family and genus placement. The taxonomic spine of the entire site.
The CC0 covers nomenclature and classification only. It does not cover descriptions or images, and we do not treat it as though it does.
Photographs, and the flowering annotations behind every phenology chart. Ground truth for our identification benchmark.
We scanned all 132 million photographs in the archive. 76.5% carry a NonCommercial licence. This site is commercial, so we cannot use them, and we do not.
The on-device identification model in the BotanAI app, and the source of the confusion data behind our lookalike pages.
We publish this model’s error rate, including the parts that are unflattering. See the accuracy benchmark.
A provenance page that lists only strengths is marketing. These are the gaps, and they are visible on the pages themselves.
Every photograph on this site carries its photographer’s name, its licence, and a link back to the observation it came from, rendered visibly under the image rather than hidden behind a tooltip. The same information is emitted as machine-readable metadata on every page.
Images licensed CC BY-SA are never cropped or composited, because a share-alike licence bites on adaptations. They render whole, letterboxed, or not at all.
If something here is wrong, we want to know, and we will say what changed and when. Write to support@botanai.app.