ByTim BiggsOctober 29, 2024 — 3.00amSaveLog in,registerorsubscribeto save articles for later.Save articles for laterAdd articles to your saved list and come back to them any time.Got itNormal text sizeLarger text sizeVery large text sizeGoogle’s search engine has been given an AI overhaul. The web giant has activated a new feature Australia-wide that will deliver answers from its generative chatbot Gemini that will sit above the usual list of web results.The feature, called AI Overviews, has been live for months in the US and elsewhere. While some users have been impressed by its ability to summarise information from multiple sources, others have expressed concerns over accuracy, potential manipulation and the interruption to advertising revenue for sites, which will be visited less often since the information is available directly on Google.Google’s AI Overviews will appear on top of other results if the system determines it will be useful.Google says the feature is designed to appear when users are making searches that would usually have them clicking through multiple links or scrolling through results. The AI-powered summaries won’t appear if regular search results are clear enough, or if they are for certain sensitive topics.The tech giant also said it was continuing to improve the AI feature, including making links to the websites it gets information from more prominent. It says AI Overviews create more opportunities for web content to be discovered, and it said users spent more time at the websites they were referred to through the feature compared to regular Google Search.“The idea is to create a jumping-off point to create a way to bring key information about a topic or a question at the top of your search results with links to go deeper, so you can quickly get to the content that makes the most sense for you,” said Hema Budaraju, Google’s senior director of product management for Search.“As we evolve the search experience with AI, we continue to stay focused on connecting people to content from a diverse range of sources. We have tappable site icons that help people get to the various citations in AI overviews, and our pre-launch testing has found this is increasing traffic to supporting websites.”At least90 per cent ofweb searchesin Australia go through Google, so it may be taking a calculated risk in changing its flagship product to compete with the likes of Microsoft’s Bing and ChatGPT. However, it has clearly seen a big jump in users moving to generative AI services where they would previously have Googled, especially in the under-24 age bracket.AI Overviews is an attempt to have it both ways. Google wants to make it clear that everything that made it the preferred search engine is still there, but that it also has quick chatbot answers.Advertisement“We add AI Overviews when we believe that it has additional value, and a higher bar of quality,” Budaraju said.“In the rare case that we make an AI Overview that does not meet the quality bar, we have a policy as well as a way of reporting feedback, that we take seriously, and then we come back to address it.”LoadingThe initial AI Overviews rollout was plagued with problems as users delighted in sharing strange and potentially dangerous results. In one instance, a user was told to putglue in their pizza. When communities on Reddit and elsewhere realised Gemini was tapping their discussions for information, and that it seemed unable to detect satire or sarcasm, some began misleading it intentionally.While Google and others are constantly shaping their AI to be better at fact-checking, they’re far from perfect. Recently, Amazon’s Alexa was caught spreading misinformation as it appeared to misunderstand information taken from a website designed to debunk conspiracy theories.Internet personality Greg Miller said on YouTube and X this month that there were 150 Planet Hollywood restaurants in Guam as an experiment to see if AI would pick it up as a fact. Google’s AI Overview was seenrepeating the informationthe next day.A significant amount of information on the internet is factually wrong, either because it’s a lie or a joke, and even more information is simply not relevant to any given web search. The danger with AI summaries is that they make their own determination about accuracy, sincerity, quality and relevance before the human user can and they present their findings earnestly.Get news and reviews on technology, gadgets and gaming in our Technology newsletter every Friday.Sign up here.SaveLog in,registerorsubscribeto save articles for later.License this articleWeb cultureAIGoogleAnalysisTim Biggsis a writer covering consumer technology, gadgets and video games.Connect viaTwitteroremail.Loading