Ninety percent accuracy is a number that travels well in product announcements. Applied to a system answering 8.5 billion queries daily, it is a catastrophe in slow motion. The Oumi analysis commissioned by The New York Times tested 4,326 searches and found that accuracy at this scale yields 600 million incorrect answers per day — a figure the Bluesky community that first flagged Google's Overview problems has now largely accepted as the frame for this conversation [19]. The argument has shifted from 'is AI Overviews getting better?' to 'does better…