Once upon a time, Google was great. They intensively monitored what people searched for, and then used that information continually to improve the engine's performance. Their big idea was that the information thus derived had a commercial value; it indicated what people were interested in and might therefore be of value to advertisers who wanted to sell them stuff. Thus was born what Shoshana Zuboff christened "surveillance capitalism", the dominant money machine of the networked world. The launch of generative AIs such as ChatGPT clearly took Google by surprise, which is odd given that the company had for years been working on the technology. The question became: how will Google respond to the threat? Now we know: it's something called AI overviews, in which an increasing number of search queries are initially answered by AI-generated responses. Users have been told that glue is useful for ensuring that cheese sticks to pizza, that they could stare at the sun for for up to 30 minutes, and that geologists suggest eating one rock per day. There's a quaint air of desperation in the publicity for this sudden pivot from search engine to answerbot. The really big question about the pivot, though, is what its systemic impact on the link economy will be. Already, the news is not great. Gartner, a market-research consultancy, for example, predicts that search engine volume will drop 25% by 2026 owing to AI chatbots and other virtual agents.