Google AI Overviews criticised for widespread ‘lies’ to users

Luke Meredith

Google’s widely used AI (artificial intelligence) summaries, which appear at the top of search results, are facing growing criticism after a report warned they are spreading incorrect information on a massive scale.

The tool, known as AI Overviews, is meant to give users quick and clear answers, but experts say its mistakes could affect millions of people every day.

The concerns come from an analysis carried out by the AI startup Oumi for The New York Times. The study found that Google’s AI summaries are correct around 91% of the time. While this sounds strong on paper, the problem becomes much larger when combined with Google’s enormous reach.

Google handles about 5 trillion searches each year, meaning even a small error rate could lead to huge numbers of people seeing wrong or misleading answers. Some researchers have described this impact as unlike anything seen before in human history.

One reason for the issue is how large AI models work. These systems are designed to respond confidently, even when they do not fully understand the topic. As a result, they can sound convincing while still being incorrect.

Studies have also shown that many users do not check these answers against other sources. This habit, known as “cognitive surrender,” means people are more likely to trust information that looks official and well-written, even if it is wrong.

There are also environmental concerns linked to the rapid growth of AI tools like Google’s summaries. AI systems depend on huge data centres that use large amounts of electricity and water. While AI can help improve efficiency in areas such as renewable energy, its fast expansion is putting extra pressure on power supplies and natural resources. If this trend continues, experts warn it could make it harder to cut emissions and control energy costs in the future.

Google and other companies point out that the technology is improving. Newer AI models are often more advanced and accurate than earlier versions. However, some researchers say these improvements bring new problems. More powerful systems may give answers that appear well-researched but are not properly backed up by reliable sources. In some cases, AI-generated citations don’t actually support the claims being made, which can confuse users even more.

These findings offer a stark reminder of the dangers of an overreliance on AI-generated answers and content. At Engage Web, we understand the value of a human-led approach to online success while AI SEO (search engine optimisation) is all the rage. If you need help getting business online, talk to our team today.

About
Latest Posts

Luke Meredith

Content Editor at Engage Web

At home in both the digital sector and the world of communications, Luke thrives on deadlines and has held roles in healthcare, retail and digital marketing, gaining a sound knowledge of a wide range of sectors. He’s a massive swot as well, holding a first-class degree in BA Communication and Media, and now uses his writing skills and keen editorial eye in his role as Content Editor here. Luke helps to oversee our clients’ content requirements, while also looking after a pool of freelance writers across three continents.