Artificial intelligence and machine learning are being applied to farm animal welfare monitoring, with potential to detect welfare problems earlier and more consistently than human observation alone.
AI welfare monitoring represents a genuine opportunity to improve farm animal welfare at scale, but its effectiveness depends critically on how it is used. Systems that detect lame cows days earlier than visual observation can improve welfare outcomes -- but only if the data triggers prompt treatment, not just recording. Welfare monitoring that generates data without action is a welfare theatre, not welfare improvement. The risk is that AI monitoring is deployed to demonstrate due diligence rather than to drive genuine change. Welfare standards that require response to welfare alerts generated by monitoring systems, combined with independent audit of response rates, would convert monitoring data into actual welfare improvements.