what's different is that the misinformation is designed to be ingested by LLMs to influence _their_ outputs, which are then presented to users as "knowledge" without necessarily attribution of sources. This is worse that the current state where you might enter a search term "morality of the Israeli army" and see a list of websites that are influenced by SEO-wrangling and paid ad word placement, but at least you can click on each website and sometimes get an idea of the bias of the website (if you read the small print) or maybe do a search about the website to see what others might say. All that is either impossible or much more difficult with AI generated content so 99% of users won't bother, but instead will go with "ChatGPT said that xxx".
