Google rolled out its new AI Overviews feature in the United States this week, and to say that the launch hasn’t gone smoothly would be an understatement. The tool is meant to augment some Google searches with an AI response to the query, thereby saving you from having to click any further. But many of the results produced range from bizarre and funny to inaccurate and outright dangerous, to the point where Google should really consider shutting down the feature.

The issue is that Google’s model works by summarizing the content from some of the top results that the search query elicits, but we all know that search results aren’t listed in order of accuracy. The most popular sites and those with the best search engine optimization (SEO) will naturally appear earlier in the rundown, whether they provide you with a good answer or not. It’s a matter for each site whether they want to give their readers accurate information, but when the AI Overview then regurgitates content that could endanger someone’s welfare, that’s on Google.