A Guardian investigation has found that Google’s AI Overviews—the generative AI summaries displayed at the top of search results—have repeatedly surfaced false or misleading medical advice, in some cases directly contradicting established clinical guidance. Health experts and charities warn that when these AI-generated summaries appear first in search results, they may be mistaken for authoritative medical advice, potentially placing people at risk of serious harm.

The Guardian investigation identified multiple cases in which Google AI Overviews surfaced health advice that experts said was inaccurate, misleading, or potentially dangerous—particularly because the summaries appeared prominently at the top of search results.

Because Google AI Overviews appear at the top of search results and present information as direct answers, health experts warned that users may assume the summaries are authoritative or clinically sound. Unlike traditional search results, which require users to evaluate multiple sources, AI Overviews often surface a single synthesized response—reducing opportunities for comparison, context, or verification. Experts warned that in health-related searches, this design can blur the line between general information and medical guidance, particularly for people seeking answers during moments of anxiety or urgency.

In one case described by experts as “really dangerous,” Google AI Overviews advised people with pancreatic cancer to avoid high-fat foods. Health specialists said this guidance was the opposite of what patients are typically advised.

Anna Jewell, director of support, research and influencing at Pancreatic Cancer UK, said the recommendation was “completely incorrect” and could jeopardize patients’ ability to tolerate treatment.

“If someone followed what the search result told them then they might not take in enough calories, struggle to put on weight, and be unable to tolerate either chemotherapy or potentially life-saving surgery,” Jewell said.

Experts explained that people with pancreatic cancer often struggle to maintain weight and require high-calorie, high-fat diets to remain strong enough for treatment. Advising patients to restrict fat intake could therefore increase the risk of poorer outcomes.

Another example involved searches for “normal liver blood test ranges.” Google AI Overviews generated summaries listing numerous numerical values but failed to account for variables such as age, sex, ethnicity, nationality, or clinical history—factors that significantly affect what is considered “normal.”

Pamela Healy, chief executive of the British Liver Trust, said the summaries were alarming and potentially misleading.

“What the Google AI Overviews say is ‘normal’ can vary drastically from what is actually considered normal,” Healy said.

She warned that people with serious liver disease—many of whom show no symptoms until later stages—could mistakenly believe their results were healthy and decide not to attend follow-up medical appointments.

“It’s dangerous because it means some people with serious liver disease may think they have a normal result then not bother to attend a follow-up healthcare meeting,” she added.

The Guardian also found incorrect information surfaced for searches related to vaginal cancer symptoms and tests. In one instance, Google AI Overviews listed a Pap test as a diagnostic test for vaginal cancer.

Health experts said this was medically wrong.

Athena Lamnisos, chief executive of The Eve Appeal, a gynecological cancer charity, said: “It isn’t a test to detect cancer, and certainly isn’t a test to detect vaginal cancer – this is completely wrong information. Getting wrong information like this could potentially lead to someone not getting vaginal cancer symptoms checked because they had a clear result at a recent cervical screening.”

She also raised concerns about inconsistency, noting that repeating the same search produced different AI summaries that pulled from different sources.

“That means that people are getting a different answer depending on when they search, and that’s not good enough. Some of the results we’ve seen are really worrying and can potentially put women in danger,” she said.

The investigation found similar issues with AI Overviews related to mental health conditions, including psychosis and eating disorders.

Stephen Buckley, head of information at mental health charity Mind, said some summaries offered advice that was “incorrect, harmful or could lead people to avoid seeking help.”

He added that AI-generated summaries can miss essential nuance and may reflect biases, stereotypes, or stigmatizing narratives, particularly when summarizing complex mental health information without sufficient context, saying they “may suggest accessing information from sites that are inappropriate.”

Health organizations including Marie Curie, Pancreatic Cancer UK, The British Liver Trust, The Eve Appeal, and mental health charity Mind emphasized that people often turn to search engines during moments of stress, fear, or crisis—conditions that increase the likelihood they will trust and act on the first information presented.

Stephanie Parker, director of digital at end-of-life charity Marie Curie, said: “People turn to the internet in moments of worry and crisis. If the information they receive is inaccurate or out of context, it can seriously harm their health.”

Concerns about inconsistency over time were raised by The Eve Appeal, which reported that repeating the same health-related search produced different AI summaries pulling from different sources. Experts said this variability means users may receive different medical guidance depending on when they search—an outcome they described as unacceptable for health information.

A Google spokesperson said that many of the health-related examples shared with the company were based on “incomplete screenshots.” From what Google could assess, the AI Overviews in question linked “to well-known, reputable sources” and encouraged users to seek expert medical advice.

The Guardian investigation also comes amid broader concern about how consumers interpret AI-generated information, particularly when it is presented in authoritative formats. In November, a separate study found AI chatbots across multiple platforms provided inaccurate financial advice, while similar concerns have been raised about AI-generated summaries of news content.

Google said the vast majority of its AI Overviews are factual and helpful, and that the company continues to make quality improvements to the feature. A spokesperson said the accuracy rate of AI Overviews is comparable to other long-standing Google Search features, such as featured snippets, which have existed for more than a decade.

The company added that when AI Overviews misinterpret web content or miss important context, Google takes action in line with its policies.

A Google spokesperson said: “We invest significantly in the quality of AI Overviews, particularly for topics like health, and the vast majority provide accurate information.”

Q: What are Google AI Overviews?
A: Google AI Overviews are generative AI summaries that appear at the top of search results, designed to provide quick answers without requiring users to click through to external websites.

Q: Why is their placement a concern for health information?
A: Because AI Overviews appear above traditional search results, users may treat them as authoritative medical advice—especially during moments of anxiety or urgency.

Q: What types of health information were affected?
A: The investigation documented errors related to cancer nutrition, liver function testing, women’s cancer screening, and mental health conditions.

Q: Did experts raise concerns about consistency?
A: Yes. Several health organizations noted that identical searches produced different AI summaries at different times, pulling from varying sources.

Q: How did Google respond?
A: Google said most AI Overviews are accurate, link to reputable sources, and are continuously improved. The company stated that when AI Overviews misinterpret content or miss context, it takes action under its policies.

Based on the documented reporting, this investigation highlights a core risk of AI-generated health summaries: prominence can be mistaken for authority. When AI Overviews appear above traditional search results, users may assume the information is vetted, stable, and clinically sound—even when it contains errors or lacks essential context.

From a journalistic perspective, the concern is not simply that AI systems can make mistakes, but that those mistakes are amplified by design. Health information is uniquely sensitive, and inaccurate or inconsistent summaries can influence decisions before users ever consult a qualified professional.

Google has said that the AI Overviews examined in the investigation linked to reputable sources and encouraged users to seek expert medical advice. That position underscores a deeper accountability question raised by the investigation.

When an AI-generated health summary appears first, presents itself as a direct answer, and contains incorrect guidance, who is responsible for the potential harm?

Health experts note that people searching for medical information are often anxious or seeking immediate clarity, and may act on what they see first rather than continuing to evaluate sources. In that context, simply advising users to “seek expert advice” may not be sufficient to offset the influence of an incorrect or misleading AI-generated answer.

As AI systems increasingly mediate access to health information, the unresolved issue is not only the quality of the sources cited, but whether responsibility lies with the user, the underlying sources, or the system that synthesized and elevated the information in the first place.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.