Google's New AI Search Is Already Spewing Misinformation
It may be Gemini season, but things aren't exactly looking good for Google.
About 26 years ago Google set about revolutionizing the web by delivering helpful and accurate information to millions of users worldwide. Today, it’s planning on doing that all over again with the help of AI.
I’m, of course, talking about Gemini and Google’s new AI search. Launched at I/O and recently rolled out to the public, Gemini has been upending your typical Google searches with AI overviews — boxes that appear at the top of the page and are intended to get you to your answer faster. Sounds useful in theory, but the results have been… less than ideal.
Google’s new AI-powered search has gotten off to a rocky start, espousing all sorts of misinformation. Apparently, you can run with scissors (it can increase your heart rate and focus), non-toxic glue is a great pizza topping, dogs did play in the NBA.
Artificial Irrelevance
For the past week, users on X have been posting examples of Google’s worst inaccuracies. SEO extraordinaire Lily Ray, for example, has been documenting some of the most egregious accounts she’s come across this week, while @Goog_Enough, an account dedicated entirely to the new AI tool and its pitfalls, has been compiling the most outlandish responses from Gemini so far. Let’s just say, if you have a medical affliction, it’s probably better to consult a doctor rather than Gemini. At this point, it seems all too keen to recommend “drinking urine” to cure just about any illness.
Behind all the wild answers is a phenomenon known in the AI world as “hallucinations,” which is when a Large Language Model like ChatGPT or Google’s Gemini entirely invents bits of information and presents the delusion as truth. As we’ve written about previously, hallucinations are still the biggest glaring problem with AI search.
It’s been entertaining to watch Gemini’s hallucinations unfold, but when considering Google’s ambitions — AI Overviews don’t currently appear for every topic, but they should become ubiquitous by the end of the year — they could also be a concern.
Google, after all, isn’t just any product, it’s the way most people access the internet and find answers to (sometimes very important) questions. And if that wasn’t enough stakes for you, Google’s AI search is being unleashed in the middle of a major U.S. election.
IS THERE A FIX?
So where do we go from here? The short answer is no one knows. Obviously, Google is intent on pushing into AI search, but without any real proof that it can make Gemini more accurate. For now, it’s more of a fingers-crossed situation.
What originally seemed like a fairly effective and straightforward idea — to short-cut and streamline the flow of information using AI — has, so far, spelled a fair amount of disaster. Without vast improvement bad of Gemini may well outweigh the good.
