Analyzing Large language models chatbots: An experimental approach using a probability test