Beyond Landmark Error: Agreement Between AI-Assisted and Manual Cephalometric Tracing at the Level of Clinical Measurements: A Systematic Review

Background: Artificial intelligence has revolutionized the field of dentistry. This systematic review evaluated the clinical performance of artificial intelligence (AI)-assisted cephalometric tracing compared with manual or expert digital tracing. Methods: We searched the PubMed/MEDLINE, Scopus, and Web of Science databases from inception to 20 May 2026. Studies directly comparing AI-assisted tracing with manual or expert digital tracing of the same lateral cephalograms and reporting at least one clinically relevant cephalometric outcome were included in the review. Two authors independently screened records. Risk of bias was assessed using an adapted QUADAS-2 framework. A quantitative meta-analysis was planned but not performed. Findings were synthesised narratively by measurement, platform, and workflow, with reporting informed by the Synthesis Without Meta-analysis (SWiM) guideline. Results: Twenty studies were included. Direct AI-versus-reference ICCs were reported in more than three studies, but their formulations and reference procedures varied. Small average differences sometimes coexisted with wide individual limits of agreement. The numerical evidence did not establish a consistent ranking of SNA, SNB, and ANB. Conclusions: AI-assisted tracing may be useful as a clinician-verified adjunct, but its performance depends on the measurement, platform, and workflow. The evidence does not establish uniform clinical interchangeability with manual or expert digital tracing.