Blended AI visibility score of 29.2 masks sharp differences by question type

An author's daily AI citation tracker recorded a blended visibility score of 29.2 on 21 October, with each answer from two web-grounded models rated 0 to 3 for how well it identifies the tracked entity. Broken down by question type, the entity performed well when asked about the author by name (88.9%) but almost never came up in open genre questions (16.7% for genre recommendations and 0% for genre, comparison and research prompts). The blended figure was driven largely by how many prompts fell into each bucket, so changing the question mix alone would move the score without any change in model behaviour. The author also cautioned that buckets of four or fewer answers are too small to support trends, and that the 12.5-point gap between the two models amounts to just two fully correct answers on a single day. The author plans to pool a week of results per bucket before drawing comparisons.