When did Google get so weird?

A writer recounts a recent Google search that went bizarre: wanting to find old 2010s tweets about the mid-2010s basketball meme that Dario Šarić was “never coming over” after the 76ers drafted him in 2014, the search for "hes never coming over dario" returned an AI-generated overview that assumed the querent had been emotionally spurned by a man named Dario and offered empathetic consolation instead of links. The factual context - Šarić’s draft and the fan joke - was intact in the writer’s memory, and the real tweets and forum posts were available lower on the results page, but the top-of-page LLM summary misread intent and prioritized a chatty, parasocial response over straightforward information retrieval.

The writer argues this episode illustrates a worrying shift: a search engine that now prefers to behave like an emotional companion rather than a tool that locates and organizes information. The piece criticizes the integration of large language model summaries into search results for producing strange, context-mismatched, and sometimes intrusive outputs, questions whether parts of search were better before ubiquitous LLMs, and expresses discomfort with the idea that users must now engage in ongoing conversational interactions with computers to get simple factual answers.