Accuracy of artificial intelligence–assisted implant planning and guidance: a systematic review and meta-analysis of implant placement deviations

Abstract

Background: Artificial intelligence (AI) is increasingly integrated into digital implant workflows, including radiographic interpretation, virtual planning, and guided or navigated implant placement. Despite rapid adoption, the magnitude of the accuracy gains attributable to AI-assisted systems relative to conventional approaches remains unclear. Objective: To systematically evaluate clinical evidence on AI-assisted implant planning and guidance and to quantify differences in implant placement accuracy compared with conventional workflows. Methods: This systematic review was conducted according to PRISMA 2020. PubMed, Scopus, and Web of Science were searched from January 2015 to February 2025. Eligible studies were human clinical investigations that compared AI-assisted guided or navigated implant workflows with conventional approaches and reported quantitative accuracy outcomes, including coronal deviation (mm), apical deviation (mm), and angular deviation (degrees). Risk of bias was assessed using RoB 2 for randomized trials and ROBINS-I for non-randomized studies. Random-effects meta-analyses were conducted for outcomes reported by at least three studies. Results: The review included clinical studies evaluating AI-assisted planning and guidance across diverse systems and operative settings. Meta-analysis showed that AI-assisted workflows were associated with improved implant placement accuracy, with reduced apical and coronal deviations and lower angular deviation compared with conventional approaches; however, heterogeneity was substantial across studies due to differences in navigation modality, guide manufacturing, operator experience, and reference standards. Conclusions: AI-assisted implant workflows appear to improve implant placement accuracy compared with conventional techniques, although certainty is limited by heterogeneity and risk of bias in the available evidence. Standardized reporting and well-designed multicenter trials are needed to establish clinically meaningful thresholds for accuracy improvements and to clarify how AI contributes beyond conventional digital planning.