I wonder if Google is going to try out "AI"-generated titles that are directly summarized from the page content by machine, treating the page title and headings as inputs to the model.
Problem solved WRT copyright issues relating to news articles. If the AI derived content (a la GitHub copilot) is deemed as original "unlicensed" content, no reason to force users to visit the website. (it's been a while since the news media and Google had their legal battles, and I'm unsure what the end resolution was then)
artificial intelligence (or at least their corporate puppet masters) are fighting for copyright law protection on insights the AI derives from reading copyrighted pages and content on the internet.