Traceable references
Connect documentary claims to material readers can inspect; do not generate fictitious citations.
We are designing a method for selecting reference material, reviewing AI-assisted output and explaining uncertainty. These are development principles, not a claim of scientific accreditation.
Generative AI can produce plausible text that is not factually supported. We aim to use targeted document retrieval and identifiable references so that significant claims can be checked against their source material.
Depending on the question, useful evidence may include peer-reviewed research, established veterinary organisations, professional guidance and applicable regulatory documentation. Commercial descriptions and individual testimonials must be labelled appropriately.
Connect documentary claims to material readers can inspect; do not generate fictitious citations.
Distinguish labelling information, an interpretation, public guidance and unsupported assumptions.
Point out missing data, uncertain inferences and circumstances requiring professional advice.
A comparison method does not become reliable simply because it produces a number. Any scoring approach needs a defined scope, permissible indicators, transparent weighting, missing-data rules and criteria for withholding a score.
Ingredient lists alone cannot universally establish nutritional suitability. Needs differ by species, life stage, physiological state and medical circumstances. This website makes no claim that a Petfood-Score rating has been scientifically validated.
From nutrition-related AI to research workflows, automation and software development, we can define a practical technical scope.