About TreeVelo
How identification works
Your photos go to a vision AI model (Claude, from Anthropic) that names the plant, lists the visible features it used and offers other possibilities. When you have TreeVelo Plus and the first answer is unsure or looks like a toxic lookalike, a second engine, Plant.id, checks it too.
The AI's own sureness is shown in words (likely, possible, not sure) because a language model rating itself is not a measured probability. Only when Plant.id provides a probability do we show a percentage, labelled as theirs.
What it can't do
- It can be wrong, especially between lookalikes, from a poor photo, or for rare plants.
- We haven't published an accuracy figure for the website. Internal tests exist, but they aren't an independent evaluation.
- It can't tell you a wild plant or mushroom is safe to eat. Don't eat anything based on an app.
- There are no human experts in TreeVelo. If a pet may have eaten a plant, call a vet or a poison service. For a tree or branch that may be dangerous, ask a qualified arborist.
Where pet-safety information comes from
A fixed table based on the ASPCA's toxic and non-toxic plant list overrides the AI for every plant it covers. Plants outside the table are marked as an unverified AI estimate. Veterinary review of the table is still pending.
Your photos and data
- Photos are resized on your device and location data is removed before upload.
- TreeVelo's server forwards them to the AI providers named above to get an answer and doesn't keep them. Those providers' own terms apply to what they receive.
- My Garden is stored in your browser. Export or delete it any time in My Garden.
- A random identifier stored in your browser counts your free identifications; "Delete my data" removes it.
- Read the full privacy policy and terms.