I don't know the inner working of chatgpt, but my best guess would be that they use something like Levenshtein distance on word to match to the nearest known wo

I don't know the inner working of chatgpt, but my best guess would be that they use something like Levenshtein distance on word to match to the nearest known word to reduce the space of known words and to accommodate to typos. When it replied to you, it took the exact extract from the input that is between quotes but internally it is mapped to "consequences" closest word to "consequence," if we consider the comma as any character.