Chemistry-transforming applications of artificial intelligence

Artificial intelligence is increasingly used across chemical research, but its practical value depends on where it intervenes in the workflow and how reliably it generalizes beyond curated benchmarks. This review adopts a pipeline-based perspective: we first decompose chemical research into stages that generate, transform, and validate information and then map AI methods onto these stages with an emphasis on measurable outcomes and failure modes. We summarize classical machine learning and emerging chemistry-informed and foundation model approaches, highlighting recurrent pitfalls such as data leakage, overfitting, and extrapolation. We then critically survey applications spanning knowledge extraction, experimental data interpretation (including data from spectroscopy, mass spectrometry, and microscopy), property and reactivity prediction, synthesis planning, and closed-loop experimentation, focusing on representative case studies that illustrate transferable design patterns rather than exhaustive coverage. Across these domains, the evidence supports a shift from static structure-property thinking toward dynamic, data-rich decision cycles where models prioritize experiments, extract mechanistic descriptors from complex signals, and guide iterative refinement under constraints. We conclude by outlining requirements for robust deployment and discussing near-term directions for chemistry-grounded foundation models and intelligent experimental platforms.