Following an investigation by “The Guardian” newspaper, “Google” has withdrawn some health summaries that rely on artificial intelligence, due to these summaries containing incorrect and misleading information, which may harm users’ health.
The company confirmed that these summaries, which use generative artificial intelligence to provide a quick overview of various topics and appear at the top of search results, are generally considered “useful” and “reliable,” but it admitted that some health-related summaries provided inaccurate information.
The investigation revealed a “serious and alarming” case where “Google” provided incorrect information about liver function tests, which could lead patients with serious liver diseases to believe that their results are normal.
“The Guardian” reported that searching for “normal range for liver blood tests” showed numbers without sufficient context or consideration of differences in gender, age, or health condition, which experts considered misleading. They also warned that what the “Overview of Liver Test Results” service classifies as normal may differ significantly from medical reality, which could lead patients to ignore necessary follow-up appointments.
As a result of this investigation, “Google” removed the “Overview of Liver Test Results” service from the results of some related searches, and confirmed that it is working on making comprehensive improvements in cases where the context is insufficient, and that it is taking the necessary actions in accordance with its policies.
Vanessa Hebditch, Director of Communications at the “British Liver Trust,” welcomed this step, but expressed concern about the possibility of misleading information appearing if questions are asked in a different way, and warned of the continued risks associated with health information produced by artificial intelligence.
The newspaper pointed out that using alternative search terms may lead to inaccurate results, raising the concerns of experts who emphasized that interpreting liver function tests is a complex process that cannot be reduced to mere numbers.
“Google,” which controls about 91% of the global search engine market, confirmed that it is reviewing the new examples provided by “The Guardian,” and affirmed that artificial intelligence summaries only appear when it is confident in the quality of the answers, and that it is continuing to review performance in various fields.
For her part, Sue Farrington, Chair of the Patients Information Forum, welcomed the removal of these summaries, but felt that this step was insufficient, stressing that maintaining public trust requires a broader addressing of the problem of the accuracy of health information provided by artificial intelligence, especially since millions of people around the world have difficulty accessing reliable health information.
