Artificial Intelligence and Metazoan Nervous Systems: From Local Processing to Valenced States of the Organism and the Inference of Suffering

Recognising animals as sentient does not establish whether an individual animal is suffering at a particular time. This structured critical narrative review examines how biological evidence can support that inference. Drawing on evolutionary and comparative neuroscience, animal-welfare research and studies of artificial systems, we examine proposed links among stimulus detection, wider neural modulation, valuation, recurrent integration and consequences for the whole organism. Artificial systems provide comparisons in which learning and adaptive processing can be implemented without independently establishing subjective experience. Their implementation neither establishes feeling nor shows that these functions are unnecessary in conscious animals. Amphioxus, zebrafish, birds, mammals and octopus illustrate distinct questions about neural organisation, internal-state regulation and valuation. Birch et al.’s dimensions guide comparison of possible forms of experience without ranking species on a single scale. Exploratory literature searches, a targeted PubMed audit and a supplementary Scopus coverage check informed source identification and verification. Evidence was appraised separately for each claim, considering alternative explanations, marker validity, dependence among measures, interventions and changes over time. The framework distinguishes capacity for experience, occurrence of a particular state and its welfare significance while assessing confidence in each inference. The proposed biological links organise this assessment; they are not a validated causal sequence or scale of consciousness. Both brief severe experiences and sustained adverse states warrant consideration. AI can extend observation, but its classifications require independent biological validation and evaluation under their intended conditions of use.