The issue was two German publishers who were accused of fraudulent practices by AI Overviews that were not supported by the underlying links, and therefore counted as Google’s «own statements,» the court found.
The publishers sent a cease and desist letter to the search engine that got ignored, resulting in the lawsuit asking for Google to remove the statements, which was upheld.
The previous interpretation was that search engines merely quoted their sources before sending the user on, and full liability for this would ruin Google’s functionality.
AI Overviews, however, generate «independent, new, and substantive statements,» the court said, which are not necessarily supported by the underlying data — and are not needed for the search engine to function.
According to the NY Times, Google’s Gemini 3 model is inaccurate around 9% of the time and includes wrong links 56% of the time, resulting in millions of wrong answers every day, Ars Technica writes.
Read more: The Decoder, Ars Technica, and Engadget.