Google has removed some of its AI Overviews health summaries after a Guardian investigation found that the summaries contained "dangerous" flaws. The removals came after the newspaper discovered that Google's generative AI feature delivered inaccurate health information at the top of search results, potentially putting seriously ill patients at risk.
Experts flagged the results as alarming, warning that the AI model's definition of "normal" often differed from actual medical standards. They noted that this could lead to patients with serious liver conditions mistakenly believing they are healthy and skipping necessary follow-up care.
The investigation found that searching for liver test norms generated raw data tables without essential context, and the AI feature failed to adjust these figures for patient demographics such as age, sex, and ethnicity. The Guardian reported that Google's page ranking algorithm has long struggled with SEO-gamed content and spam, feeding these unreliable results to its AI model.
Despite removing some summaries, other potentially harmful answers remain accessible. Experts have warned that typing slight variations of the original queries into Google still prompts AI Overviews, making it easy for readers to miss that the numbers might not be accurate for their test.
Google has maintained that it invests in the quality of AI Overviews, particularly for health topics, and that "the vast majority provide accurate information." However, the company declined to comment on why some summaries remained active after The Guardian's investigation.
Experts flagged the results as alarming, warning that the AI model's definition of "normal" often differed from actual medical standards. They noted that this could lead to patients with serious liver conditions mistakenly believing they are healthy and skipping necessary follow-up care.
The investigation found that searching for liver test norms generated raw data tables without essential context, and the AI feature failed to adjust these figures for patient demographics such as age, sex, and ethnicity. The Guardian reported that Google's page ranking algorithm has long struggled with SEO-gamed content and spam, feeding these unreliable results to its AI model.
Despite removing some summaries, other potentially harmful answers remain accessible. Experts have warned that typing slight variations of the original queries into Google still prompts AI Overviews, making it easy for readers to miss that the numbers might not be accurate for their test.
Google has maintained that it invests in the quality of AI Overviews, particularly for health topics, and that "the vast majority provide accurate information." However, the company declined to comment on why some summaries remained active after The Guardian's investigation.
just saw this news and i'm like totally floored 
google's ai overviews are supposed to give us health info but apparently they're delivering info that's straight outta badland 
did u know that the AI model's definition of 'normal' is actually way off from actual medical standards? 
it's crazy how they just spit out raw data without context or adjusting for things like age, sex, and ethnicity! 
it's like google's AI model is playing doctor and we all know how that ends 
but what about the ones that aren't? 
u can still get unreliable results by just tweaking the original query 
and a whopping 70% of those users trust AI-generated answers? 
. I'm totally worried about people getting misinformation online, especially when it comes to serious health issues like liver conditions 
. It's crazy that experts have been warning Google about this for ages, and now we're seeing the consequences 
. Can't believe these AI models are being fed all sorts of SEO-gamed content too... just another example of how much tech can screw up our lives 
its not good, esp 4 those wit liver probs 
. It's crazy that they didn't catch these flaws sooner. You gotta feel for those patients who are super sick and might've made decisions based on info that's just plain wrong 
. And let's not forget, they're already removing some of the problematic ones, so that's a win in my book 
!