Artificial intelligence has progressed from a specialized computational aid to a prominent organizing concept across target identification, molecular screening, lead optimization, synthesis planning, drug repurposing, and development strategy. However, the visibility of the field has grown faster than the evidentiary frameworks needed to distinguish methodological novelty from pharmaceutical value. This bibliometric and thematic review examines how the research landscape should be interpreted after the most promotional phase of the AI drug-discovery discourse. The review combines a transparent dual-database corpus-construction protocol with performance analysis, science mapping, and structured thematic coding. Rather than treating publication growth, citation impact, collaboration density, or benchmark performance as direct evidence of maturity, the synthesis evaluates what each indicator can establish and where additional validation is required. The resulting interpretation separates methodological expansion from evidence maturation and distinguishes computational capability, comparative benchmark performance, external generalization, prospective experimental confirmation, developability, clinical evaluation, and routine pharmaceutical readiness. The review further proposes claim-level hype-cycle indicators based on validation realism, reproducibility, data suitability, evidence provenance, prospective confirmation, and correspondence between model endpoints and pharmaceutical decisions. Bibliometric structures are interpreted as properties of the constructed corpus rather than universal representations of the field, while thematic clusters are treated as organizing devices rather than proof of scientific coherence or translational success. Important limitations include database-dependent coverage, incomplete visibility of proprietary industrial research, terminology drift, uneven reporting of negative findings, and the difficulty of attributing program-level outcomes to individual computational components. The central implication is that the next phase of AI-enabled drug discovery should be evaluated through decision-relevant evidence, traceable data practices, realistic validation, and explicit boundaries between prediction and pharmaceutical utility.