Google's AI-powered search summaries, known as 'overviews', are meant to give quick answers to questions, but users say they're often dangerously wrong. From recommending high-fat foods to pancreatic cancer patients to misreporting liver function results, these automated summaries have prompted alarm among those relying on them for serious health information. Reddit users shared examples showing that people with little medical knowledge could be misled, risking their health.
Many say the AI pulls together data from multiple sources without understanding context, sometimes inventing entirely new claims that contradict credible guidance.
Google AI Overviews: Can They Be Trusted?
The AI overviews are designed to give users a quick summary at the top of Google search results. They attempt to condense information from multiple web sources into a concise answer, providing a shortcut for busy users. While the feature can save time, experts warn it often mixes facts and errors, leading to misinformation.
For instance, one AI overview reportedly told pancreatic cancer patients to avoid high-fat foods, even though dietary fat can be important for some patients' survival. Another incorrectly suggested normal results for critical liver tests, which could make seriously ill people think they are healthy. The AI uses common patterns across sources rather than evaluating accuracy, which can result in confidently stated but entirely false information.
Many Reddit users reported encountering incorrect AI summaries in everyday searches. One said, 'Even with very little knowledge of the thing I am searching, many times I have noticed the AI overview is wrong'.
'In a rush, I stupidly trusted it to tell me which terminal I needed to be dropped off at and almost missed my return flight home after the holidays... I will never trust it again lol,' said a user who noted that it caused her delays in traveling.
Some noticed obviously false historical examples, like AI claiming Elizabeth I had been pregnant: 'I recently wrote it off entirely for anything even a little serious because it utilized Elizabeth I (aka the virgin queen) as an example of pregnant monarchs'.
Misleading Health Advice Hits Vulnerable Users
A worrying pattern emerged in health-related searches. Users with serious conditions, like neurological diseases, reported banding together to campaign for fixes after receiving incorrect AI guidance.
'I have a neurological disease and the patient community actually banded together to campaign for Google to fix the 'AI overview' for our disease, as it was so false and harmful. Why do we have to do that when Google is a rich tech giant with all the resources to get this right?'
Even seemingly minor queries could go wrong. A Redditor revealed that when they had food poisoning, Google AI Overview gave them information on what to eat and drink. However, they noted that they found 'an article on HealthLine that said everything BUT what it was recommending was ok to drink. I went and got an IV in the end, but it was alarming'.
Nonsense Answers From Google AI Overview
The AI often mixes sources together, sometimes creating bizarre results. Users described being told to 'smell with my eyes' or encountering nonsense answers about video game lengths. 'Google AI overview once advised me to "smell with my eyes". It's a joke. That's what happens when AI mashes multiple sources together in one jumbled overview with no actual intelligence'.
Some users have found workarounds, like adding "-ai" to Google searches to remove the AI summary, though this isn't always effective.
Analysts stress that Google's AI overviews lack the nuanced judgement of humans, particularly in areas like health or safety. The system counts common answers rather than verifying accuracy, which can amplify errors, especially when people trust AI blindly.
'It doesn't consider the source; it simply looks at the most common answer(s) across all sources. Since there's a lot of misinformation, it very often is more common than the actual good sources. Garbage in, garbage out'.
Reddit users warned that over-reliance could have serious consequences: 'It's wrong ALL the time, and it's bad enough that it's telling people they've got scary rare diseases that they don't have - that alone can cause harm'.
Many Reddit users have switched to alternative search engines or adjusted queries to avoid the AI. 'No kidding. AI is dangerous & I wish people would stop using it... It should discourage everyone, but people would rather use it than take 10 mins to compose a simple text or email'.
