Artificial intelligence is reshaping expectations in analytical chemistry, but many of its challenges are not new. Chemometricians have grappled for decades with issues of data quality, overfitting, and validation - precisely the pitfalls now facing AI. This presentation will contrast the empirical discipline of chemometrics with the current wave of generative and predictive AI, identifying lessons that can ensure AI contributes meaningfully to chemical analysis rather than adding more “smoke and mirrors.” Through real-world examples, I will explore how AI tools can assist, rather than replace, human reasoning, and how rigorous validation must remain the foundation of trustworthy chemical insight.
