The search giant made AI Overviews — which are displayed atop a search results page before traditional links to the Web — available to all U.S. users in May after spending one year trialing a limited earlier version.

The feature was widely panned after screenshots of factually inaccurate answers circulated across the internet, such as a pizza recipe that listed glue as an ingredient and an answer claiming that former U.S. President Barack Obama is Muslim.

Google acknowledged the “odd and erroneous overviews” and announced updates to the product in a blog post in late May. These updates added restrictions to which queries would display AI answers and curbed user-generated content from websites like Reddit from serving as source material for answers.

“I have enough evidence to say that quality is only improving,” Hema Budaraju, a senior director of product at Google told Reuters in an interview on Wednesday. She pointed to data Google collects internally, which showed that users with access to the feature reported higher levels of satisfaction and searched for longer queries than users who did not.

AI Overviews is now coming to the Brazil, India, Indonesia, Japan, Mexico and Britain, in local languages such as Hindi and Portuguese.

Google is also adding hyperlinks to the feature. Websites will be displayed to the right side of the AI-generated answer. The company is also internally testing a further update that would add links directly within the text of the overview.

The updates come amid concerns voiced by the media industry about the possibility of losing out on referral traffic from consumers who clicked through to publishers’ websites. Budaraju said the new update would have a “three-way benefit” for Google, consumers and publishers.

Last week a U.S. judge ruled Google had an illegal monopoly on search, clearing the way for a trial that could force the breakup of Alphabet. AI advances from rivals like Microsoft-backed OpenAI could pose an even bigger threat.

—Kenrick Cai, Reuters