We tightened sentiment analysis accuracy for AI answers.
Sentiment is now derived from the sentence directly surrounding the brand mention, preventing other brands' tone from leaking into your signal. Korean negation patterns ("not recommended", "not good", "no issues") are recognized whether they appear before or after the keyword.
A distinct neutral signal (amber ⚬) makes the positive·neutral·negative tri-state immediately readable. When a brand is cited in an objective context it is honestly shown as neutral, and answers that don't mention the brand at all are unified to neutral as well — boosting measurement consistency.