Last week, Google launched its AI search results for millions of users, aiming to deliver a better search experience. However, instead of enhancing the experience, the AI produced some bizarre and inaccurate results, such as advising users to put glue on their pizza to help the cheese stick and to eat rocks.
Google acted swiftly to remove these inaccurate AI results, termed AI Overviews, but the damage and meme-ification had already occurred. In a blog post published on Thursday by Google’s head of search, Liz Reid, the tech giant attributed the inaccurate results to “data voids”, odd user queries, and emphasized that despite the issues, AI results are leading to “higher satisfaction” with Search. Reid argued that AI Overviews generally don’t “hallucinate”; they sometimes misinterpret existing web data.
“There’s nothing quite like having millions of people using the feature with many novel searches,” Reid writes. “We’ve also seen nonsensical new searches, seemingly aimed at producing erroneous results.” She also noted that a “very large number of faked screenshots” of AI Overviews have been spreading online. It’s worth mentioning that some queries, like “what mammal has the most bones?”, are legitimate, especially from inquisitive children. Additionally, Google automatically included millions of people in this feature, leading to significant backlash and articles explaining how to disable it.
Reid’s blog also details how Google is fixing AI Overviews by limiting their appearance for “nonsensical” queries and satirical content. This measure addresses the issue of well-known satirical sites like The Onion and Reddit users like “fucksmith” not being filtered out initially. Furthermore, Reid compares AI Overviews to another longstanding Search feature, featured snippets, which highlight information from relevant web pages without using generative AI. According to Reid, the “accuracy rate” for featured snippets is “on par” with AI Overviews.
Ultimately, the introduction of AI Overviews has been another PR blunder for Google. Meanwhile, the company is in a race to compete against OpenAI and AI search startups like Perplexity, rumoured to be worth $3 billion. If Google aims to stay competitive, it must move quickly but also maintain user trust, which may be challenging to regain after AI Overviews suggested eating Elmer’s glue.
