<sec> <title>UNSTRUCTURED</title> Generic sentiment analysis tools are widely deployed in digital mental health and longitudinal text research to infer psychological change. Lexicon-based approaches (eg, NRC), supervised emotion classifiers (eg, GoEmotions), and single-shot large language model (LLM) prompts typically operationalize affect as the frequency or probability of valenced tokens at the message level. While effective for detecting overt affective shifts, the limits of this operationalization remain underexamined. This Viewpoint presents a single-subject longitudinal corpus (n=531 reflective messages across six months) to illustrate a construct–measurement misalignment. The subject reported a substantial psychological transformation characterized by reframing, integration, and narrative restructuring rather than shifts in emotional tone. Phase A (lexicon counts), Phase B (supervised emotion probabilities), and Phase C (LLM affect ratings) were applied under standardized aggregation schemes (per-message scoring, arithmetic mean, binning). Across methods, no consistent longitudinal trend emerged in affective indices. We argue that this null result is not a failure of AI per se, but a measurement blind spot: when transformation occurs at the level of narrative meaning rather than valence frequency, generic sentiment tools may remain insensitive. We propose a construct-specific analytic framework emphasizing alignment between psychological target constructs and computational operationalization. Implications are discussed for responsible deployment of AI in longitudinal digital health and narrative analysis contexts. </sec>