Skip to content
LeafRanger
Never written by AI. Not at any tier.

“Not listed as toxic” is not the same as safe

Nearly every plant app and listicle collapses pet safety into two buckets: toxic, or fine. There is a third, it is large, and it is the one that matters — plants nobody has assessed for your animal. We keep it separate and we say so.

Assessed — harmful
Assessed — cleared
Not assessed

The third lamp never lights. That’s the point.

Three states, not two

Every plant lands in one of three compartments

The distinction that matters is between the second and the third. Most sources — and most apps built on them — merge those two, and that merge is where people get hurt.

affects

Assessed, and harmful

Someone checked this plant against this animal and found it causes harm. The app names the animals specifically — a plant can be a problem for cats and not for dogs.

safeFor

Assessed, and cleared

Someone checked and explicitly cleared it for that animal. This is a positive finding, held separately — it is not inferred from the absence of a warning.

neither list

Nobody has checked

No assessment exists for that animal. The app says exactly that. It does not round the gap down to “probably fine”, because that is a guess about your pet.

The rule underneath all of it: an empty toxicity record renders as “we don’t know”. Never as “non-toxic”. There is no condition under which the app converts an absence of information into reassurance.
The dangerous shortcut

Absence of a warning is not a clearance

If an app only stores “which animals is this harmful to”, then every plant nobody ever tested looks identical to a plant that was tested and passed. One of those is safe for your cat. The other is simply unstudied — and there are far more of them than people assume.

  • Assessed and harmful is a finding. Someone looked.
  • Assessed and cleared is also a finding. Someone looked, and the answer was no harm. We store this separately, on purpose.
  • Outside both is not a finding at all. It is the absence of one, and it is the largest region of the three.
harmfulclearedeverything else — unassessed

An app that stores only the left circle has to treat the whole white field as if it were the right one.

Why one answer isn’t enough

The same plant can be lethal to one animal and fine for another

This is not an edge case, and it is the reason a single “toxic: yes/no” flag can’t do the job. Lilies are the example everyone in a veterinary practice knows.

One plant · lilies
CatsAssessed — seriously harmful

Among the most dangerous houseplants there is for a cat. If you keep cats, this is a plant to not have in the house at all, rather than one to place carefully.

DogsAssessed — cleared

A positive finding, not an absence of one. That distinction only survives because the record stores “cleared for” separately from “harmful to”.

Collapse those two halves into one flag and you have to pick: warn every dog owner unnecessarily, or reassure every cat owner dangerously. We don’t pick — the app asks which animals you actually keep, and answers for those.

The one rule with no exceptions

No AI has ever written a toxicity record, and none ever will

Everywhere else in this product, a missing field gets researched and filled in automatically — that is a feature we’re proud of. This field is carved out of it entirely.

verified

A checked dataset, or a person

Those are the only two things permitted to write this field. Both leave a record of who decided and on what basis.

The model — refused, at every tier

Even when the field is empty. Especially when the field is empty, because that is exactly when a language model is most willing to produce something plausible.

Why this one and not the others: a wrong care tip produces a sad plant, and you find out slowly enough to fix it. A wrong toxicity answer sends someone’s cat to a vet — or, worse, fails to. Those two mistakes do not deserve the same safeguards, so they don’t get them.
Practical placement

You don’t always have to get rid of it

A plant that’s harmful to your animal isn’t automatically a plant you can’t own — with one honest exception. What matters is whether the animal can actually reach it, and cats make that a much harder question than dogs.

Toddler can reach
Dog can reach
Cat can reach
  • A closed door works on every animal on this list. Height works on some of them.
  • Cats knock things down before they eat them — a heavy pot on a narrow ledge is its own hazard.
  • Fallen leaves and trimmings are the usual route in. Sweep up after pruning.
  • The water in a saucer counts too, for plants whose sap is the problem.
  • If an animal is a determined chewer, rehome the plant rather than negotiating with it.

The exception: for the small number of plants that are severely toxic to a species you keep — lilies and cats being the one everyone should know — placement isn’t the answer. Don’t have it in the house.

If it has already happened

Ring a vet. Don’t look it up first.

We are a plant app. We are not going to give you a treatment, a home remedy, or a list of symptoms to check against — every one of those things costs time, and some of them do real harm. This is the whole of our advice:

  1. Ring a vet nowYour own practice, or an out-of-hours emergency service. Do this first, before searching anything.
  2. Take the plant with youOr a clear photo of it, and a piece of the leaf. Identifying it correctly changes the treatment.
  3. Tell them when, and how muchRoughly what was eaten and how long ago. Both matter more than the symptoms you can see.

An animal that seems fine can still be in trouble, and with some plants the damage happens well before anything looks wrong. That’s the reason this page won’t give you symptoms to weigh up — it isn’t a judgement you should be making from a website.

Asked constantly

Pet safety, answered plainly

Is my plant safe for my cat?
That depends on the plant, and the honest answer is sometimes “nobody has checked”. In the app, each plant's record tells you one of three things for each animal you keep: assessed and harmful, assessed and cleared, or not assessed. We don't collapse the last two together.
Why won't you just give me a list of pet-safe plants?
Because a list like that is only as honest as its worst entry, and most of them quietly include plants nobody ever assessed. Ours would have to be a much shorter list than anyone else publishes, and it would still be the wrong shape — what you need is the answer for the plant in front of you, and the animals you actually keep.
What does “not assessed” actually mean?
That no checked source records a finding for that animal. Not that it's dangerous, and not that it's fine. It's the absence of information, and it's a much larger category than most people realise — so we show it as its own state rather than folding it into “non-toxic”.
Could the AI be wrong about toxicity?
The AI never writes it. Toxicity is the one field in the entire product that a model is forbidden to fill in — at any tier, and especially when the field is empty, which is exactly when a language model is most willing to produce something plausible. Only a checked dataset or a person can set it.
My plant is toxic to my dog. Do I have to get rid of it?
Usually not — it depends whether the animal can actually reach it, and a closed door does more than a high shelf. The exception is the small number of plants that are severely toxic to a species you keep, lilies and cats being the one everyone should know: those shouldn't be in the house at all.
My pet has eaten some of a plant. What do I do?
Ring a vet now, and take the plant or a clear photo of it with you. We're not going to offer a remedy or a list of symptoms to check — both cost time, and this isn't a judgement to make from a website.
Does it cover rabbits, birds and other pets?
Coverage varies a lot by animal, and that's precisely why the three-state model matters: for a less-studied animal you'll see “not assessed” far more often. That's a real answer about the state of the evidence, not a gap we've papered over.
Is a plant that's toxic to pets also toxic to people?
Not necessarily — they're recorded separately, because the answer genuinely differs. The app holds both, and where one has been assessed and the other hasn't, it says so rather than assuming they match.

Check the plants you actually own

Tell the app which animals live with you, and every plant in your garden carries its verdict for each of them — including, where it applies, the honest “nobody has assessed this”.