When Sundar Pichai, Google’s chief executive, unveiled a generative artificial intelligence feature for the company’s search engine last month, he and his colleagues demonstrated the new capability with six text queries that the public could try out.
Questions included “how to clean a fabric sofa” and “what should I use to remove a coffee stain from my carpet.” They aimed to highlight how Google’s new feature, AI Overviews, can generate comprehensive and useful summaries of information over and above traditional search results.
But as of Friday, only one in six queries still yielded an AI review, according to tests by The New York Times. Instead, the feature was noticeably less prevalent. A search for “what should I use to remove coffee stains from my carpet” now returned a snippet of text from a website, JDog Carpet Cleaning & Floor Care, while “how to clean a fabric sofa” was replaced with a link to the website of HGTV has the answer. (Search results may vary by user and location.)
The disappearance of AI Overviews for some of the searches appears to have been part of a wider pushback after the new technology produced a host of falsehoods and errors – including recommending glue as part of a pizza recipe and suggesting that people swallow rocks for nutrients. Users complained loudly on social media about the errors, in many cases outright mocking Google.
Liz Reed, who was recently promoted to head Google’s search division, wrote in a blog post Thursday that the company has scaled back AI reviews in certain ways, launching “additional trigger refinements” to offer more thoughtful responses about health, disabling misleading advice and limiting the inclusion of satire and user responses from forums such as Reddit.
“We will continue to improve when and how we show AI Overviews and strengthen our protections,” she wrote, adding that Google is working on updates to improve a wide range of search results.
Ashley Thompson, a Google spokeswoman, said in a statement Friday that the company had made more than a dozen technical updates to its systems.
“AI Overviews helps people with a large number of Search queries today by serving as a point of reference to content on the web.” The company added that while it is making adjustments to improve AI Overviews, it is not retiring the feature long-term .
The retreat was a blow to Google’s efforts to keep up with rivals Microsoft and OpenAI, the maker of the ChatGPT chatbot, in the frenzied race to be the AI leader. It also highlighted the difficult strategic choice Google faces about whether to embrace AI technology that may not be reliable or keep its hugely popular search engine the same — and risk falling behind its competitors.
Google chose to go slower than Microsoft, which put more conversational AI into its Bing search engine early last year. Google, which has significantly more users than Bing, tested AI features for its search engine a year before introducing AI Overviews. The company said the new feature will roll out immediately to users in the United States and to more than one billion people by the end of the year.
But in the end, Google “should have rolled this out more slowly,” said Patrick Hall, assistant professor of decision sciences at George Washington University’s School of Business. “Once something like this happens, you really have to step back. And if nothing else, there is damage to the company’s reputation.
Google, the leader in Internet search for more than two decades, has been struggling since OpenAI launched ChatGPT in 2022. Some people in the tech industry believe that the chatbot’s ability to generate answers is a serious threat to Google’s search engine, which is the most – the popular way to get information online.
Since then, Google has worked aggressively to regain its AI edge, releasing a family of technologies called Gemini, including new AI models for developers. The company has also implemented the technology in YouTube, Gmail and Docs, helping users create videos, emails and drafts with less effort.
Last month, Ms. Reid said at Google’s developer conference that the search engine would do more than Google for users with AI Overviews. She highlighted the increasingly complex queries that Google can answer with the feature, but those capabilities have not yet rolled out for users.
On stage, simpler questions like “how do I get the campfire smell out of my clothes” yielded AI answers, including airing, adding baking soda and spritzing with lemon juice.
But when users got their hands on the new service, they discovered that AI Overviews sometimes generated wrong — or downright dangerous — answers, including recommending putting non-toxic glue on the pizza to make the cheese stick. He also misunderstood some of the websites he cited and got the presidential story wrong.
Of the six questions Google released to the public, the only one that consistently triggered an AI review on Friday was “what are some interesting science projects I can do with my 12-year-old son.”
The answer? Growing crystals, extracting DNA from saliva and writing a message with invisible ink.
