The application of artificial intelligence in the field of dental restoration for designing dental crowns: a meta-analysis Authors: Jiaxuan Hu, Binding Bai, Yuzhou Wang, Bo Hu, Xiaoming Fu Published: 2026 ABSTRACT BackgroundThis meta-analysis aimed to investigate the differences between dental crowns designed by artificial intelligence (AI) and those designed through traditional computer-aided design (CAD) processes to provide a reference for the clinical application of AI in the field of dental crown restoration.MethodsSearches were conducted in PubMed, Web of Science, CENTRAL, EMBASE, WFPD, VIP, and CNKI to collect in vitro studies on AI-designed dental crowns and traditional CAD-based dental crowns. After the data were extracted and the risk of bias was assessed with the Quality Assessment Tool for In Vitro Studies (QUIN Tool), a meta-analysis was conducted to clarify the differences between AI-designed dental crowns and traditional CAD-based dental crowns. The search period ranged from the establishment of each database to January 2026. Two researchers independently screened the articles, extracted their basic information and assessed the risk of bias. Between-study heterogeneity was assessed using Cochrane Q and I² statistics.ResultsA total of 12 articles were ultimately included. The meta-analysis revealed that the design time of the AI group was shorter than that of the traditional CAD group (mean difference, -289.03; 95% confidence interval, -450.18 to -127.88; P = 0.0004), and the occlusal accuracy of the former was better than that of the traditional CAD group (mean difference, -67.39; 95% confidence interval, -132.36 to -2.42; P = 0.04); however, no significant difference was observed between the occlusal morphology discrepancy (mean difference, -0.02; 95% confidence interval, -0.06 to 0.02; P = 0.34), and the cusp angle in the AI group was smaller than that in the traditional CAD group (mean difference, 3.31; 95% confidence interval, 1.12 to 5.50; P = 0.003).ConclusionsCompared with traditional CAD, AI significantly reduces the required design time while producing similar occlusal morphology results and demonstrating greater stability in occlusal accuracy. However, compared with that of the traditional CAD method, the cusp angle designed by AI still needs further optimization. Overall, AI design shows promise as an auxiliary tool for dental crown design, but further algorithm optimization and high-quality clinical trials are needed to confirm its long-term clinical performance.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12903-026-08680-1. Background Tooth defects, as some of the most common clinical oral diseases, refer to damage or varying degrees of abnormalities on the surfaces or in the internal hard tissues of teeth. The defects not only directly affect the facial appearances and physiological functions (such as chewing) of patients but also have negative effects on the quality of life and mental health of patients [1, 2]. In the field of prosthodontics, dental crowns, as a common fixed restoration method for tooth defects, have undergone an evolution process concerning their manufacturing technology from traditional artificial manufacturing to digital manufacturing. The traditional artificial manufacturing techniques (lost-wax casting) have dominated for a long time. However, the techniques have significant limitations, such as being time-consuming, labor-intensive, and overly reliant on the operational experience and technical proficiency of skilled workers. Due to these limitations, it is difficult to meet the higher demands of modern stomatology regarding the precision, functionality and aesthetic effects of restorations. With the advancement of digital technology, computer-aided design/computer-aided manufacturing (CAD/CAM) technology has gradually become the mainstream method for producing dental crowns. Digital manufacturing technology supported by CAD/CAM has achieved a high degree of automation and precise control in the dental crown production process, significantly improved the accuracy and efficiency of dental crown production, shortened the production cycle, and reduced the reliance of the procedure on the experience of technicians [3]. Although CAD/CAM technology offers distinct advantages over the lost-wax casting process, in traditional CAD/CAM workflows, the crown design phase relies primarily on the use of dental CAD software by technicians. This approach still requires significant manpower and time, particularly in complex cases where technicians must devote substantial hours to making meticulous adjustments. AI refers to the ability of algorithms encoded in information technology to learn from data. Its advantage lies in its ability to simulate human thinking, learn independently, solve problems, and achieve automation [4]. As an emerging technology, if AI is applied to dental crown design, algorithms can be used to simulate the thought processes of technicians, and automated design-based software programs can be developed to further save time and effort, improve the efficiency of design, and achieve automated and precise dental crown design. With the development of AI technology, many scholars have focused on the application of AI in the design of dental crowns, but their conclusions vary. Chau et al. [5] reported that when AI was used to design single-crown restorations for molars, the morphological differences were smaller than those of natural teeth, which demonstrates the feasibility of using AI in a single-crown design scenario. Feng et al. [6] demonstrated that AI design not only provides enhanced efficiency but also automatically and accurately reconstructs the three-dimensional (3D) shapes of missing maxillary anterior tooth crowns. However, some studies have shown [7–9] that during the design process, AI may encounter dynamic interference, inaccurate automatic digital edge detection results, and the loss of some feature information during the 3D shape restoration procedure applied to the target dental crown. These issues require technicians to make further adjustments and modifications; otherwise, doctors will need to spend a considerable amount of time on chairside adjustments, and the final clinical performance of the dental crown may even be affected. Therefore, it is important to conduct a meta-analysis to evaluate the differences between AI-designed dental crowns and traditional CAD-based dental crowns to provide a reference for the application of AI in the field of dental crown restoration. Methods This study adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and the protocol was registered at the international prospective register of systematic reviews PROSPERO (registration: CRD420261352582). Search strategy From the establishment of each database to January 2026, a literature retrieval procedure was conducted with respect to the application of AI technology for fabricating dental crown restorations in multiple databases, such as PubMed, Web of Science, Cochrane Central Register of Controlled Trials (CENTRAL), EMBASE, Chinese Medicine Premier’s Wanfang database (WFPD), VIP Chinese Science and Technology Journal Databases (VIP), and China National Knowledge Infrastructure (CNKI), were searched according to a combination of subject terms and free terms. The search scope covered relevant studies on AI-designed dental crowns and traditional CAD-based dental crowns. The search terms included “artificial intelligence”, “intelligence”, “computational intelligence”, “crown”, “crown appearance”, “crown morphology”, and “crown prostheses”. The search retrieval formula was as follows: ((((crown [Title/Abstract]) OR (crown appearance [Title/Abstract])) OR (crown morphology [Title/Abstract])) OR (crown prostheses [Title/Abstract])) AND (((artificial intelligence [Title/Abstract]) OR (intelligence [Title/Abstract])) OR (computational intelligence [Title/Abstract]). Inclusion and exclusion criteria The inclusion criteria were as follows: (1) research subjects: dental crowns; (2) research types: in vitro studies of AI-designed dental crowns and traditional CAD-based dental crowns; (3) publication languages: Chinese and English; (4) publication time: from the establishment of each electronic database to January 2026; and (5) outcome measures: design time, occlusal morphology discrepancy, occlusal accuracy and cusp angle. The exclusion criteria were as follows: (1) reviews or animal experiments; (2) studies for which detailed data or the full text could not be obtained; (3) studies with obvious abnormalities in the relevant data, from which no further information could be obtained by contacting the authors; (4) studies with incomplete data and those from which the original data could not be extracted; and (5) studies lacking a control group. Study selection and data extraction Two researchers independently conducted the study screening and exclusion steps and ultimately determined the included studies through discussions and comparisons. If differences of opinion were encountered during the screening process and if no consensus could be reached through discussion, a third researcher arbitrated the decision. The extracted data included the name of the first author, year of publication, study type, study object, sample size, and outcome measures. Quality assessment The methodological quality and risk of bias of the trials included in the review were assessed by two reviewers using the QUIN Tool, with disagreements resolved by a third reviewer. The QUIN Tool comprises 12 specific evaluation criteria (detailed in Table 2). To assess the risk of bias of in vitro studies using this tool, each criterion was evaluated against four rating categories: adequately specified, inadequately specified, not specified, and not applicable. A standardized scoring system was applied for quantitative evaluation: each criterion was rated as adequately specified = 2 points, inadequately specified = 1 point, not specified = 0 points, and not applicable=exclude criteria from calculation. The scores were then added to obtain a total score for a particular in vitro study. The final score was calculated using the formula: Final score = (Total score×100) / (2×number of criteria applicable). According to the percentage value obtained, each study was classified into one of three bias risk levels: low risk of bias for scores > 70%, moderate risk of bias for scores ranging from 50% to 70%, and high risk of bias for scores < 50% [10]. Statistical analysis Two researchers were responsible for exporting and processing the data. The articles were recorded in detail using NoteExpress software, and a meta-analysis was conducted with Review Manager 5.3 software. In the meta-analysis, we calculated overall continuous variables estimates using the mean differences (MD) and the upper and lower limits of the 95% confidence intervals (CI) to assess overall efficacy from all the eligible studies. Heterogeneity was assessed by Q statistic (P < 0.10 indicating significant heterogeneity), and I squared (I2) statistic. An I2 value of more than 50% represented high heterogeneity; thus, the randomeffects model was adopted. An I2 less than 50% represented low heterogeneity, in which a fixed-effects models were used. Statistical significance was declared if the P-value was < 0.05. Results Data selection In all, 576 articles were retrieved from the databases employed in this study, including 60 from PubMed, 147 from the Web of Science, 368 from EMBASE, and 1 from CNKI. Moreover, 23 articles were retrieved manually. On this basis, 24 duplicate articles were identified and eliminated, and 575 articles were initially screened. The researchers read the titles and abstracts of these articles, eliminated 552 articles on the basis of the principle of relevance, and then rescreened them to obtain 23 articles. Ultimately, based on the preset inclusion and exclusion criteria for the included studies, a detailed full-text study of these 23 articles was conducted, and 12 of them that satisfied the imposed requirements were identified for a subsequent quality risk assessment. The screening process employed in this study is shown in Fig. 1. Fig. 1The flow chart of the study screening process The flow chart of the study screening process Characteristics of the included studies This study ultimately included 12 articles, the basic information of which is summarized in Table 1. Table 1Basic characteristics of the included studiesFirst AuthorYearStudy ObjectSample SizeOutcome MeasuresECChen Y [11]2022Dental crown1212Occlusal morphology discrepancy, Cusp angleKollmuss M [12]2013Dental crown3939Occlusal morphology discrepancyKollmuss M [13]2016Dental crown2222Occlusal morphology discrepancyLitzenburger AP [14]2013Dental crown525Occlusal morphology discrepancyChao JH [15]2024Dental crown3030Cusp angleDing H [16]2023Dental crown1212Cusp angleLiu CM [17]2024Dental crown1515Design timeLei L [18]2024Dental crown5858Design timeEnder A [19]2011Dental crown1212Design timeKoudai Nagata [20]2025Dental crown1010Design time, Occlusal accuracyLiu CM [21]2025Dental crown1010Occlusal accuracyWu ZQ [22]2025Dental crown3333Design time, Occlusal accuracyAbbreviations: E Experimental group, C Control group Table 1Basic characteristics of the included studiesFirst AuthorYearStudy ObjectSample SizeOutcome MeasuresECChen Y [11]2022Dental crown1212Occlusal morphology discrepancy, Cusp angleKollmuss M [12]2013Dental crown3939Occlusal morphology discrepancyKollmuss M [13]2016Dental crown2222Occlusal morphology discrepancyLitzenburger AP [14]2013Dental crown525Occlusal morphology discrepancyChao JH [15]2024Dental crown3030Cusp angleDing H [16]2023Dental crown1212Cusp angleLiu CM [17]2024Dental crown1515Design timeLei L [18]2024Dental crown5858Design timeEnder A [19]2011Dental crown1212Design timeKoudai Nagata [20]2025Dental crown1010Design time, Occlusal accuracyLiu CM [21]2025Dental crown1010Occlusal accuracyWu ZQ [22]2025Dental crown3333Design time, Occlusal accuracyAbbreviations: E Experimental group, C Control group Basic characteristics of the included studies Abbreviations: E Experimental group, C Control group Risk of bias Six included studies had a low risk of bias, six had a medium risk, and none had a high risk (Table 2). Table 2Quality assessment results of in vitro studies using QUIN ToolStudyC1C2C3C4C5C6C7C8C9C10C11C12Total ScoreRisk of Bias Calculation(%)Risk of Bias InterpretationChen Y [11]2222220220222083.33Low-riskKollmuss M [12]2122221100221770.83Low-riskKollmuss M [13]2122212211222083.33Low-riskLitzenburger AP [14]2122211211221979.17Low-riskChao JH [15]2122210101221666.67Medium-riskDing H [16]2122210200221666.67Medium-riskLiu CM [17]2112210200221562.5Medium-riskLei L [18]2122212200221875Low-riskEnder A [19]2122212200221875Low-riskKoudai Nagata [20]2122210200221666.67Medium-riskLiu CM [21]2122200200221562.5Medium-riskWu ZQ [22]2122210200221666.67Medium-riskScoring: 2, Adequately Specified; 1, Inadequately Specified; 0, Not Specified; NA, Not ApplicableC1 Clearly stated aims/objectives, C2 Detailed explanation of sample size calculation, C3, Detailed explanation of sampling technique, C4 Details of comparison group, C5 Detailed explanation of methodology, C6 Operator details, C7 Randomization, C8 Method of measurement of outcome, C9 Outcome assessor details, C10 Blinding, C11 Statistical analysis, C12 Presentation of results Table 2Quality assessment results of in vitro studies using QUIN ToolStudyC1C2C3C4C5C6C7C8C9C10C11C12Total ScoreRisk of Bias Calculation(%)Risk of Bias InterpretationChen Y [11]2222220220222083.33Low-riskKollmuss M [12]2122221100221770.83Low-riskKollmuss M [13]2122212211222083.33Low-riskLitzenburger AP [14]2122211211221979.17Low-riskChao JH [15]2122210101221666.67Medium-riskDing H [16]2122210200221666.67Medium-riskLiu CM [17]2112210200221562.5Medium-riskLei L [18]2122212200221875Low-riskEnder A [19]2122212200221875Low-riskKoudai Nagata [20]2122210200221666.67Medium-riskLiu CM [21]2122200200221562.5Medium-riskWu ZQ [22]2122210200221666.67Medium-riskScoring: 2, Adequately Specified; 1, Inadequately Specified; 0, Not Specified; NA, Not ApplicableC1 Clearly stated aims/objectives, C2 Detailed explanation of sample size calculation, C3, Detailed explanation of sampling technique, C4 Details of comparison group, C5 Detailed explanation of methodology, C6 Operator details, C7 Randomization, C8 Method of measurement of outcome, C9 Outcome assessor details, C10 Blinding, C11 Statistical analysis, C12 Presentation of results Quality assessment results of in vitro studies using QUIN Tool Scoring: 2, Adequately Specified; 1, Inadequately Specified; 0, Not Specified; NA, Not Applicable C1 Clearly stated aims/objectives, C2 Detailed explanation of sample size calculation, C3, Detailed explanation of sampling technique, C4 Details of comparison group, C5 Detailed explanation of methodology, C6 Operator details, C7 Randomization, C8 Method of measurement of outcome, C9 Outcome assessor details, C10 Blinding, C11 Statistical analysis, C12 Presentation of results Analysis results of the outcome measures Design time The 5 included studies compared the time required for designing dental crowns between the AI group and the traditional CAD group [17–20, 22]. The heterogeneity test results (I2=99%) indicated heterogeneity among the studies, and thus a random-effects model was used for the analysis. The results revealed that the design time of the AI group was significantly shorter than that of the traditional CAD group [MD=-289.03, 95% CI (-450.18, -127.88), P=0.0004] (Fig. 2). Fig. 2Forest plot comparing the design time of the AI-designed dental crowns (AI) and traditional CAD-based dental crowns (TD) Forest plot comparing the design time of the AI-designed dental crowns (AI) and traditional CAD-based dental crowns (TD) Occlusal morphology discrepancy The 4 included studies analyzed the occlusal morphology discrepancy of dental crowns designed by AI and those designed by the traditional CAD method [11–14]. A heterogeneity test (I2=85%) indicated heterogeneity among the studies; thus, a random-effects model was used for analysis. The results revealed that there was no statistically significant difference between the occlusal morphology discrepancy of the two groups [MD=-0.02, 95% CI (-0.06, 0.02), P=0.34] (Fig. 3). Fig. 3Forest plot comparing the occlusal morphology discrepancy of the AI-designed dental crowns (AI) and traditional CAD-based dental crowns (TD) Forest plot comparing the occlusal morphology discrepancy of the AI-designed dental crowns (AI) and traditional CAD-based dental crowns (TD) Occlusal accuracy The 3 included studies compared the accuracy of the occlusal surfaces of dental crowns designed by AI and those designed by the traditional CAD method [20–22]. A heterogeneity test (I2=97%) indicated heterogeneity among the studies; thus, a random-effects model was used for analysis. The results revealed that the occlusal accuracy of the AI-designed dental crowns was better than that of the traditional CAD-based crowns [MD=-67.39, 95% CI (-132.36, -2.42), P=0.04] (Fig. 4). Fig. 4Forest plot comparing the occlusal accuracy of the AI-designed dental crowns (AI) and traditional CAD-based dental crowns (TD) Forest plot comparing the occlusal accuracy of the AI-designed dental crowns (AI) and traditional CAD-based dental crowns (TD) Cusp angle The 3 included studies analyzed the cusp angles of dental crowns designed by AI and those designed by the traditional CAD method [11, 15, 16]. A heterogeneity test (I2=0%) indicated extremely low heterogeneity among the studies; therefore, a fixed-effects model was used for analysis. The results revealed that the cusp angles designed by AI were smaller than those designed by the traditional CAD method [MD=3.31, 95% CI (1.12, 5.50), P=0.003] (Fig. 5). Fig. 5Forest plot comparing the cusp angles of the AI-designed dental crowns (AI) and traditional CAD-based dental crowns (TD) Forest plot comparing the cusp angles of the AI-designed dental crowns (AI) and traditional CAD-based dental crowns (TD) Discussion Crown restoration is a widely used treatment method in the field of stomatology. The advent of CAD/CAM technology has greatly promoted technological innovation in this field. It has rapidly replaced traditional techniques such as lost-wax casting and ceramic stacking, gradually transforming cumbersome manual operations into efficient digital and mechanized process flows. Although CAD/CAM technology has many advantages, it still requires considerable manpower, material resources and time for dental crown design tasks. In addition, traditional CAD technology relies on manual design procedures, and the dental crowns manufactured through this approach still have certain limitations in terms of personalized effects. Clinically, problems such as dental crowns with poor marginal fits and deviations in the designs of the occlusal and adjacent surfaces often occur, which usually require additional adjustments or re-fabrication [23]. With the rapid development of computer technology and the advent of the big data era, the applications of AI design in dental crown design scenarios have gradually increased, injecting new vitality into CAD/CAM technology. By integrating AI algorithms into CAD/CAM systems and leveraging advanced deep learning (DL) algorithms and big data analyses, it is possible to automatically design dental crowns that are more accurate and that highly simulate the shapes of natural teeth. This innovation has further enhanced the personalized customization levels of dental crowns. A total of 12 in vitro studies were included in this review. The experimental group and the control group were AI-designed dental crowns and traditional CAD-based dental crowns, respectively. A meta-analysis was conducted on the design time, occlusal morphology discrepancy, occlusal accuracy, and cusp angle of AI-designed and traditional CAD-based dental crowns. This study first analyzed the time difference between AI-designed and traditional CAD-based dental crowns. The results showed that AI design had a significant advantage in terms of temporal efficiency. Compared with traditional manual labor, the advantages of AI mainly lie in its ultrahigh-speed data processing capabilities, algorithm-driven stability, continuous and efficient hardware design process, and autonomous learning and evolution capabilities of deep neural networks [24]. The traditional CAD/CAM-based dental crown restoration and manufacturing process mainly relies on manual design, which usually takes a long time [25]. However, through automated and intelligent design techniques, AI design significantly shortens the design and adjustment time, yields improved work efficiency, and reduces the demand for manpower. In addition to design time, the restoration contour design is equally important, as it directly affects the aesthetics and functionality of the restoration. The results of the present study were analyzed according to two indicators: occlusal morphology discrepancy and occlusal accuracy. Occlusal morphology discrepancy was quantified via volumetric and profile deviations of occlusal morphology, aiming to compare the surface deviation and repeatability of 3D anatomical forms between AI-designed and traditional CAD-based crowns [14]. Four early-published included studies on occlusal morphology discrepancy adopted the volume/area method [11–14]. The volume/area method is mainly used for analyzing the morphologies of 3D structures. It does not require the setting of reference points, has no related selection bias, makes morphological comparisons more standardized, and has a smaller impact on extreme values. The results revealed no significant difference between the occlusal morphology discrepancy of the AI-designed and traditional CAD-based methods. Occlusal accuracy is mainly used to evaluate the degree of 3D matching between crowns designed by AI or traditional CAD and the ideal reference model (natural teeth/standard database) [21]. Three newly published studies included in the meta-analysis of the occlusal accuracy adopted the 3D alignment fitting method [19–21]. Through fitting software such as Geomagic or Match3D, 3D alignment fitting was performed on the 3D images of the two objects via best-fit alignment. This automatic alignment method is widely used to align complex and irregularly shaped objects. The results suggested that compared with the traditional CAD method, AI design achieves greater occlusal accuracy. This might be related to the use of DL algorithms in AI technology. DL, as an important branch of AI, is widely used in various clinical tasks for automatically decision-making. DL models are constructed through artificial neural networks. Through multilayer convolutional neural networks and generative adversarial networks, datasets can be classified and learned at multiple levels. DL also automatically extracts the learning features contained in the input data to improve the quality and accuracy of the output [26]. Farook et al. [27] employed a 3D convolutional neural network model to conduct deep learning and an analysis on the 3D preparation data obtained through digital intraoral scanning technology, which achieved personalized designs for some crowns with an accuracy of up to 83%. DL algorithms can automatically extract the features of tooth morphologies by learning from a large amount of data and can conduct multilevel optimization during the design process, thereby improving the accuracy of dental crown morphologies. This fully demonstrates the potential and application prospects of DL algorithms in the field of dental restoration. Although AI has certain advantages in terms of the design of the morphologies of the occlusal surfaces of dental crowns, this study revealed that the cusp angles of AI-designed dental crowns differ from those of crowns designed by the traditional CAD method. The cusp angle refers to the angle between the cusp plane and the cusp of the longitudinal axis of a tooth. When a dental crown is designed, the cusp angle is among the key indicators for evaluating the quality of the design. It not only affects the aesthetic effect of the dental crown but is also related to its anti-fracture ability. When most dental crowns experience catastrophic overall fractures, such fractures are clinically difficult to repair. Scholars believe that a cusp angle range of 50° to 70° is clinically acceptable [28–30]. This study revealed that the cusp angles designed by AI are slightly larger than those designed by the traditional CAD method. However, the cusp angles designed by traditional CAD are closer to those of the original teeth. This might be related to the following factors. When AI design addresses high-dimensional objects (such as 3D datasets), the vector representation method induces generalization errors. The vertical-dimensional data obtained by intraoral scanners lack systematic consistency, resulting in errors that affect mainly the determination of vertical positions and randomness [31]. This results in AI-designed dental crowns having larger cusp angles than those of the original teeth, reflecting reduced the steepness of the functional cusps. In contrast, although the traditional CAD method relies on the experience of technicians and the selection of CAD software templates, it can flexibly adjust cusp angles according to the actual situation of the patient's teeth, avoiding errors caused by algorithmic and data issues in AI design scenarios. Furthermore, AI design relies on a large amount of training data. However, the sample sizes used in the existing studies are far from those required for the development of clinically applicable AI, and if the employed data sources are biased, the quality of the results will vary. This might also be one of the reasons for the deficiencies of this technique in the detailed design of the cusp angles of dental crowns. Simultaneously, several limitations of the present study should be acknowledged. First, AI, which is a new technique in dental crown design, lacks associated primary studies, as we found only 12 articles that could be included in our study. Second, we noticed that there were certain degrees of heterogeneity among some of the studies. Heterogeneity is associated with the presence of confounding factors within and across the included studies, such as tooth position, design software, and the experience or proficiency of dental technicians. Third, we did not evaluate their publication biases or perform sensitivity analyses since the number of included studies was limited and the included studies provided insufficient information. Therefore, this study is limited by its reliance on in vitro studies, which may not fully recapitulate intraoral conditions. Substantial heterogeneity across outcomes compromised the interpretability and generalizability of pooled estimates; thus, the results of this study should be interpreted as directional trends, with nonsignificant findings considered inconclusive. Considering the limitations described previously, future research should incorporate additional evaluation indicators, such as occlusal contact or the absolute marginal discrepancy, and more rigorous controlled trials should be conducted. These measures will help assess the advantages and disadvantages of AI design in the field of dental crown restoration to ultimately develop a standardized evaluation framework for AI applications in this field. Conclusion In summary, compared with the traditional CAD method, AI design can significantly reduce the time needed for design, generate similar occlusal morphology discrepancy, and perform more stably in terms of the occlusal accuracy. However, there is still room for improvement in the traditional CAD method regarding the design of the cusp angles of dental crowns. Overall, AI is suitable as an auxiliary tool for dental crown design tasks and is recommended for clinical promotion. However, it is still necessary to continuously optimize the algorithmic architectures and model performance of AI to enhance its applicability and service efficiency in the field of dental crown restoration and to provide technical support for digital dental restoration scenarios. Supplementary Information Supplementary Material 1. Supplementary Material 1.