Background <p>Intraoral scanners (IOS) facilitate dental treatment, but the efficacy in full-arch scanning remains controversial. The aim of this study was to compare arch deformations between conventional impressions (CIs) and digital impressions (DIs) across six distinct spans in the maxillary and mandibular models, using the absolute linear deviation method.</p> Methods <p>Standard maxillary and mandibular models, each with seven cylindrical landmarks added, were used as the reference. CIs and DIs as test scans (<i>n</i> = 15 each) were performed on the models using silicone impression material and three IOSs: CS3600, Trios3, and Trios5. The trueness of the distances and angles between the remaining cylinders and initial scanning cylinder were evaluated. Data were analyzed using the Kruskal-Wallis and One-way ANOVA tests, with the Bonferroni test for post hoc analysis (α = 0.05).</p> Results <p>Deviations of DIs increased gradually from smaller spans to full-arch spans, while deviations of CIs remained stable. Within a 5-tooth-units, DIs provided superior trueness compared to CIs (<i>P</i> &lt; 0.05), except for ΔL8, where the results from four impression methods were comparable (<i>P</i> = 0.28). For other measurements, CIs exhibited significantly better trueness than three IOSs (<i>P</i> &lt; 0.05).</p> Conclusions <p>The current accuracy of IOSs was insufficient for full-arch applications, but suitable for short scan ranges (fixed prostheses within a 5-unit span).</p>