It may not be impressive, unless you consider "how" the answer is derived. When you search Google, you'll receive information that pertains to the question you have asked. Google searches an index of some kind and provides links to the most relevant sites it can find.

When you ask ChatGPT a question, it generates a response by referencing the model that was used to train the algorithm. It is true that it's predicting the next tokens in the sequence, but it also demonstrates at an understanding of the world that search engines do not exhibit. This is why, at this point, large language models are subject to fantasy, they get math wrong a lot of the time, and only get a "C" on some university level physics questions. It is also why, in the near future, they will improve and surpass human abilities on most tasks.