0
I’ve tried asking Claude and ChatGPT the same question, with the same wording, and sometimes the answers disagree on basic details—not just tone or formatting. Is that mostly because the models were trained differently, or can settings and the wording of the prompt account for much of it?
I’m especially interested in how people check a claim when two fluent responses conflict. Do you compare sources, ask each model to explain its reasoning, or take a different approach? I’d like to understand how much of this is model-specific behavior and how much comes from the prompt itself.
