Ahaha, it’s… not actually reading the article. That is a hallucinated summary of what it thinks an article is most likely to say given the url, which contains the title of the article.
Ah, that does actually make sense, because ChatGPT can’t reach out to URLs right? It has its own time contained dataset, correct? Wonder if it would have a cache of that URL if the URL existed at the snapshot/time of its dataset?