What Puppy AI does, what it deliberately refuses to do, and exactly when it tells you to stop reading and call a vet or a certified trainer.
Puppy AI is a daily log with an AI coach on top of it. It helps you track potty, meals, sleep and behaviour, spot patterns in your own data, and get plain answers to routine puppy questions at any hour.
Puppy AI is not a vet, not a replacement for a certified trainer, and not a diagnostic tool. It does not diagnose illness, does not prescribe treatment, and does not take on serious behaviour cases. Where those lines are is written out below, not buried in the terms.
The rule is boring on purpose: anything that could be medical, and anything involving real fear or aggression, goes to a human. No hedging, no "it might just be…", no attempt to be helpful past the point where being helpful is the wrong instinct.
The full list of symptoms and behaviours that trigger a hard stop is on the safety and red-flag page.
Puppy AI's guidance is force-free and reward-based. It does not recommend, and will not help you use, aversive tools or techniques — no shock or e-collars, no prong collars, no alpha rolls, no scruffing, no dominance framing. This is not a stylistic preference; it is the position of the major veterinary behaviour bodies, and the evidence behind it is not close.
Where behaviour advice touches health, the guidance defers to your vet rather than competing with them.
Puppy AI is built on a large language model, and large language models are confidently wrong sometimes. The mitigations are real but partial: the model answers with your actual log in view rather than in the abstract, medical and serious-behaviour questions are routed out rather than answered, and the scope above is enforced rather than aspirational.
What that adds up to: treat it as a well-informed, tireless assistant that has read your puppy's diary — not as an authority. When something feels wrong, it probably is, and the vet is the right call. The app will tell you the same thing.