Surely it just reasoned that you made a typo and "autocorrected" your riddle. Isn't this what a human would do? Though to be fair, a human would ask you again to make sure they heard you correctly. But it would be kind of annoying if you had to verify every typo when using an LLM.