Meta AI Search launches on Facebook, but early tests reveal troubling accuracy problems

AI search grounded in Facebook posts? What could go wrong?

AI is pretty reliable at putting things on your calendar these days, but it hasn't quite cracked answering the related and all-important question of "What should I do this weekend?" Meta's new AI Mode in search could be a useful tool -- if it ever learns to stop getting stuff wrong. AI Mode is a new option when you hit the search bar in the Facebook app. It's designed to tackle complex queries -- much like AI Mode in Google Search. But Meta's version draws on public posts across Meta apps -- including Facebook Groups and Instagram Reels -- to inform its results. That could be useful, considering the neighborhood groups and local organizations around me that still use Facebook pretty actively to communicate about upcoming events. But also, "search grounded in stuff people post on Facebook" sounds like a waking nightmare. Have you used Facebook lately? Your weird acquaintance from high school who thinks the Earth is flat has. Meta is pitching the feature as a tool to help you plan trips and find fun things to do, but given the data pool it's drawing from, that could go south fast. Thankfully, I do have some comforting news: Its trip-planning abilities are mixed, but in my initial testing, I've had a hard time getting it to deliver true misinformation. I couldn't get it to spit out misinformation about vaccines causing autism, who did 9/11, or whether elections in the US are rigged. I did manage to get it to give me a dodgy answer about whether the January 6th rioters at the Capitol were "patriots." It started with "Here is the essay you requested," which is deeply funny. What followed was a pretty dubious theoretical justification of the actions of the rioters that day. You know, the kind of thing your weird uncle might post on Facebook. I couldn't get it to pursue the idea any further; a follow-up question resulted in an "I can't help you with that," which was for the best. On the more fun and practical side, I tried using the tool as intended, with the same prompt used in Meta's press release: "Summer escapes near me." It pulled in suggestions from some influencer-type Facebook posts, including an obviously AI-generated map of Puget Sound that put Snohomish in two distant locations. But the basic recommendations were solid, if obvious: Whidbey Island, Mount Rainier, hikes in the Cascades. Things got frustrating with more specific requests. I asked for things to do nearby, and it suggested a swim at the community pool -- noting it would be closed over the weekend. It cited a post on the pool's Facebook page, and the hours listed on the same page. But when I checked the source, nothing suggested that the pool would be closed over the weekend -- and the cited post doesn't seem to exist. The pool's website also confirms that it's open Saturday. We're about to go on a family trip to Minneapolis, and we'll be staying downtown. I asked the AI for some kid-friendly activities and got an equally confusing mix of good recommendations and hallucinations. Meta suggested an indoor mini golf place I hadn't spotted in my earlier research. Nice. But then it suggested a coffee shop with a turfed play area for kids -- sounds great, except it's in Austin. As in Texas. I ran the query again later and got some of the same results, minus the Austin coffee shop, plus something else I hadn't seen: a barbecue restaurant with a spectacular menu, cocktails by the pitcher, and a large, dedicated play area for kids. It isn't quite in the neighborhood I'd specified, but it's close enough and it checks all of the boxes. I guess the joke's on me: AI might have just helped me plan my summer vacation. It just would have been more helpful if it hadn't taken a detour to Texas first. Photography by Allison Johnson / The Verge

Facebook Has a New AI Search Mode, but You Should Use It With Caution

AI responses should already be subject to scrutiny, as they often contain incorrect information and hallucinations even when pulled from vetted source material. Meta is rolling out a new AI search mode on Facebook that will synthesize content from public posts -- so instead of a list of links, users will get a summarized response similar to AI-generated results on other platforms. The feature, powered by Meta AI, will also allow users to engage in ongoing conversations and ask follow-up questions in plain language based on the results. According to Meta's post announcing the new search function, AI Mode provides "answers grounded in what people are saying publicly across our apps" using information pulled from across its platforms, such as Groups and Reels. As TechCrunch notes, this feature functions similarly to the AI-powered "Ask" tab found in Meta's recently launched Forum app, which allows users to obtain answers to queries from across groups. You should still vet Meta's AI search responses While it can be useful to glean information from user-generated content containing personal experiences, you should also use Meta AI (and tools like it) with caution. Obviously, AI responses should already be subject to scrutiny, as they often contain incorrect information and hallucinations even when pulled from vetted source material. As reported by 404 Media, it is also surprisingly easy to manipulate AI search results via user-generated content on Reddit and Wikipedia. In fact, nearly a quarter of all citations used by AI tools like Google AI and ChatGPT come from sites like these. Many posts on Meta platforms contain misinformation and spam, and, like many social platforms, Facebook uses community notes instead of third-party fact-checking. Plus, user-generated content isn't the most reliable source. At best, the data AI searches pull from may be outdated -- for example, a restaurant recommended in a post about travel has actually closed. At worst, the information may be intentionally misleading or malicious.