Remember when people started asking AI tools for cooking advice, and it wound up telling them to do things like use glue to get cheese to stick on pizza? Well, people are apparently relying on that same technology to guide them through cooking this year’s Thanksgiving dinner. In fact, so many are doing so that Bloomberg reports it’s putting a real dent in the views of recipe writers who usually see traffic spike this time of year.
The problem is effectively the same one that led to Google previously recommending that people eat one rock per day: AI Overviews in Search. They provide users with a quick panel that pulls out all of the “relevant information” without requiring them to click through to a website and scroll through the admittedly annoying 2,000-word personal essay that precedes every recipe ever posted online.
This creates two issues. The first is for the recipe authors, who have put actual work—from their collected knowledge of food to the effort of prep work to the trial and error to get the final product just right—into the recipes they share. They’re getting their traffic siphoned off by the AI Overviews. Creators that Bloomberg spoke with said their traffic was down between 40% to 80% this year from previous Thanksgivings. That’s in line with the experience of other sites, too, which have reported as much as 80% declines in click-throughs since AI Overviews became more prominent.
The second problem is for people making the recipes, because there is a very real chance that they are getting bad information. Here’s the thing about AI summaries of anything: it doesn’t actually understand what it is reading. All it can do is spit back what it thinks is relevant. That’s kind of a big deal for cooking, where little errors can ruin a dish. For instance, Bloomberg talked to one cook who has a popular Christmas cake recipe. On the creator’s page for the recipe, it suggests baking it at 160°C (that’s 320°F) for an hour and a half. An AI-summarized version of that recipe recommends you bake it for three to four hours—more than twice as long. You don’t have to know a whole lot about baking to know that’s not going to turn out great.
AI-generated recipes have become a whole micro-industry. If you hop on any social platform and go looking for ideas of what to cook, there’s a good chance you’ll land on a page that looks like your standard cooking inspiration fare—but you might notice that the recipes just aren’t quite right. Best-case scenario, you’ll probably end up with a relatively bland but perfectly fine dish. Worst case, you might end up burning down your house because somewhere in the black hole that is a large language model, it decided that you should put your tinfoil-wrapped fish in the microwave on high.
Maybe grab one of those old cookbooks off the shelf this holiday season just to be safe.
Read the full article here
