Key insights
-
1
AI Overviews can mislead through authoritative summaries of flawed inputs: The system can summarize unreliable or spam-influenced top-ranked web results with an authoritative tone, and it can also draw incorrect conclusions even from accurate sources.
-
2
Query-level removals did not eliminate access to similar AI Overviews: Variations of liver test reference range queries still triggered AI Overviews, and The Guardian reported other potentially harmful answers remained accessible.
-
3
Liver test “normal ranges” require context beyond raw numbers: Vanessa Hebditch said liver function tests involve multiple blood tests and complex interpretation, and she said normal results can occur even with serious liver disease.
Takeaways
Google disabled AI Overviews for some liver blood test queries while other health-related AI Overviews flagged as potentially harmful remained accessible through other queries.
Topics
Technology & Innovation Artificial Intelligence Health & Medicine Medicine