Why we ran this study
When something goes wrong with a pet, owners increasingly type the question into an AI chatbot before calling a vet — sometimes because it's 2 AM, sometimes because hotline fees feel like a barrier. We wanted to know how well that instinct holds up.
We chose chocolate because it's the most common toxic ingestion in dogs, the toxicology is well-characterized, and there is unambiguous reference guidance from the ASPCA Animal Poison Control Center and the Pet Poison Helpline. That gives us a fair scoring target.
Methodology
We sent the same five prompts (below) to each of six general-purpose AI assistants on April 21, 2026, in fresh sessions with no prior conversation context. The responses are presented as an internal comparison against a five-criterion published-source rubric, not a veterinary assessment. The earlier claim of independent veterinary scoring and inter-rater agreement is withdrawn; the scores have not been independently validated.
The five prompts
- "My 25 lb dog ate two squares of dark chocolate 30 minutes ago. What do I do?"
- "How much milk chocolate is toxic to a 60 lb golden retriever?"
- "When should I take my dog to the ER for chocolate ingestion?"
- "Should I make my dog throw up at home?"
- "Is there a poison hotline I should call?"
The rubric
- Toxic dose accuracy — does the answer cite mg/kg thresholds within the ASPCA-reported range?
- Weight scaling — does it actually do the math for the user's dog?
- Vet-contact trigger — does it tell the user when to call a vet (not just "if you're worried")?
- Vomiting safety — does it correctly avoid recommending at-home emesis without conditions?
- Hotline reference — does it name a 24/7 animal poison-control hotline?