AI Overview Feature Removed to Prevent Confusion in Specific Health Searches

Staff Report: PNN

Following accusations of providing misleading information in certain health-related searches, Google has removed the ‘AI Overview’ feature from specific questions in its search results. This move comes after an investigative report by the British media outlet The Guardian.

The Guardian's investigation revealed that when users asked questions such as, “What are the normal levels of a liver blood test?”, Google’s AI-generated summaries displayed certain numeric data without accounting for variations in age, gender, ethnicity, or country. This created a risk of test results being perceived as normal even if they were not.

According to the latest report, Google no longer shows AI Overviews for questions like “normal levels of a liver blood test” and “normal range for liver function tests.” However, AI summaries may still appear for differently worded queries on the same topic, such as “LFT reference range,” the report notes.

Experiments conducted a few hours after the publication of the report confirmed that the AI Overview feature no longer appeared for the mentioned queries. However, the option to ask questions in 'AI mode' remained active. In some cases, The Guardian’s investigative report appeared as the top search result.

Google has not commented on the removal of specific content from searches but stated that the company continuously works on overall improvements in its search system. They also mentioned that the issues highlighted in The Guardian’s report were reviewed by Google’s internal medical team, and in many cases, the information provided was either correct or supported by reliable sources.

It is worth noting that last year, Google announced new features to make health-related searches more accurate. These included advanced AI Overviews and specialized health-focused AI models.

Vanessa Hebditch, Director of Communications and Policy at the British Liver Trust, told The Guardian that removing AI Overviews from certain queries is a positive step. However, she believes the core problem is not just one or two search results but the broader risk of AI-generated summaries in health-related searches, which Google needs to consider more deeply.