Artificial intelligence in dentistry: insights and expectations from Swiss dental professionals

Abstract

Background

The goal of this study was to explore Swiss dentists’ opinions on artificial intelligence (AI) and illustrate possible correlations to sex, age or professional background.

Methods

An online questionnaire was designed and sent to 1121 Swiss dentists by e-mail. It included questions about current feelings, hopes and worries regarding the future of AI in dentistry and enquired habitual and professional use of AI tools.

Results

After initial screening, 114 returned questionnaires were included in the final analysis of gathered data. This study revealed that 21.9% of respondents reported using AI in dentistry at least once a week. No significant differences were found between male and female participants regarding their perceptions of AI safety and utility (p = 0.823); however, a significant negative correlation was found between participants’ age and their belief in AI’s utility (p = 0.049). The belief that AI might replace jobs in the future correlated with lower perceived AI utility.

Conclusions

The findings provide insight into AI’s role in Swiss dentistry, highlighting areas for future research. Greater emphasis on digital medicine and AI in dental education is encouraged to advance the field and enhance oral health-related quality of life.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12911-025-03066-9.

Background

Switzerland’s multicultural society, with four national languages and a strong tradition of direct democracy and federalism, fosters political stability and broad participation. Its robust education system and global engagement further support innovation, including in digital technologies [1, 2]. In recent years, Switzerland has also advanced significantly in digitalization, with artificial intelligence (AI) playing an increasingly important role across sectors, including healthcare [3].

AI’s integration into dentistry offers notable opportunities, but also raises ethical concerns. Recent research emphasizes the importance of transparency and explainability in clinical applications. The European AI Act (AIA), for instance, mandates that AI systems must support clear, interpretable decision-making (Article 13) [4, 5]. Furthermore, Articles 14 and 29 highlight the need for human oversight to mitigate potential risks to patient safety and rights—concerns echoed in current dental literature [6].

The term “artificial intelligence” was coined by John McCarthy in 1956, building on ideas from Alan Turing’s earlier work on machine cognition [7, 8]. Today, modern language models like ChatGPT [9] already meet Turing’s criteria for AI, producing coherent professional-level text.

Global advances in AI research are driven by increasing computational power and vast digital data. Dental education is beginning to reflect this shift. A recent global study by Uribe et al. found that 64% of dental educators recognized the potential of large language models in teaching, with 31% already using them. Interestingly, respondents from Africa, Asia, and the Americas perceived greater AI potential than those in Europe [10].

These developments have increased AI’s role in medical research [11]. Although the quality of the impact on everyday clinical practice is the subject of controversial debate among experts, the increasing use of AI in both research and clinical settings is undeniable [12, 13]. AI is an efficient technology for evaluating and exploiting large amounts of data. In pharmacology, for example, artificial intelligence is already widely used to drive the discovery of new active ingredients [14–18].

Moreover, AI can be used in drug administration to optimize the selection of drugs for patients and predict harmful interactions. It can also improve treatment protocols by calculating adequate dosage and method of administration. Across the healthcare sector, AI can be applied in numerous ways to improve a wide range of processes [19–22].

Especially in radiological and photographic imaging, AI has great potential. Deep Learning, a sub-category of machine learning, can be used for image evaluation and analysis. Large amounts of data can be processed to find repeatable patterns or trends that can be used to predict the output. This makes it especially useful for face, object and speech recognition [23, 24]. These possibilities also apply to medical radiology. Research is already exploring chronological analysis, including transitions and progressions, beyond single radiographs [25]. AI can also support clinicians directly. It can be used for therapeutic decision-making [26]. Dentistry is no exception. Various studies show that AI-powered tools can potentially improve dental care, from diagnosis, treatment planning to risk assessment and promotes personalized, predictive, preventive and participatory dentistry [27–32].

In terms of prior research, similar studies to this one have been conducted in Germany, Austria and Switzerland (DACH countries), as well as in Croatia, Australia, and Brazil [33–36]. The study conducted by Fitzek et al. in DACH countries targeted German-speaking medical and dental students. It found that only 18.2% of participants reported to having received formal training in AI. “Significant positive correlations were found between self-reported tech-savviness [sic] and AI familiarity (r = 0.67) and between confidence in finding reliable AI information and positive attitudes toward AI (r = 0.72). […] Dental students exhibited slightly more positive attitudes toward the integration of AI into their future practices” [33]. A study by Ivanišević et al. focusing on Croatian general dentists revealed that, while 76.0% of respondents did not use AI daily, 71.0% believed that these technologies could enhance patient care. Nevertheless, the mean knowledge level of AI was assessed as relatively low [34]. Similarly, Hegde et al. conducted a study among Australian dental students and dentists, reporting that 69% of respondents believed AI would be beneficial in clinical tasks. Even though a considerable number of respondents also expressed apprehension regarding the potential misuse of AI technologies [35]. A Brazilian study by Pauwels et al. investigated the effects of an AI-focused lecture on dental students and dentists. The study found a generally positive attitude toward AI; however, one-third of participants expressed concerns regarding the technology. “After the lecture, agreement regarding the different roles of AI in oral radiology increased, overall excitement regarding AI increased, and concerns regarding the potential replacement of oral radiologists decreased” [36]. Overall, these studies show the significant influence of education on attitudes toward AI and highlight the need for better-informed opinions within dentistry and healthcare. As AI continues to take root in healthcare, many concerns and unanswered questions persist. A lack of both basic knowledge and advanced training in the field of digital medicine remains evident [37]. Consequently, many experts have called for improved IT education for healthcare professionals, including dentists [38].

This study focuses on the opinions of Swiss dentists regarding AI. It aims to assess their perceptions, thoughts, and ideas about the present and future of AI in dentistry, as well as their habitual and professional use of AI technology. By investigating these perspectives, we aimed to provide a clearer understanding of how AI is perceived within the Swiss dental community, identify potential concerns or misconceptions, and highlight areas where educational interventions might be beneficial. Insights gained from this research may assist in tailoring AI applications to better meet the needs and expectations of dental professionals, while also contributing to the ongoing discourse regarding the ethical, practical, and educational implications of AI in dentistry.

Methods

An original 20-item web-based questionnaire was designed [Supplementary file 1 Questionnaire] and posted to 1121 Swiss dentists using “SurveyMonkey”. A confidential list of 65 e-mail addresses of military dentists was requested from the Swiss military. Additionally, publicly available e-mail addresses from Swiss dentists were gathered by searching on “Google Maps”. The search included various types of clinics such as private, joint and university clinics. A total of 1056 individual e-mail addresses from all 26 cantons (Swiss member states) were identified this way. The questionnaire was distributed via a link included in an e-mail. The e-mail included the survey duration, investigator identity, and study purpose. Participation in the survey was voluntary, and no incentives were offered. The e-mail text was in German or French, depending on the language used by the respective website from which the e-mail address was obtained. Italian-speaking dentists also received e-mails in German. German, as the most widely spoken language in Switzerland, was used for the final survey. No additional advertising was conducted. SurveyMonkey used several cookies, most of which expired within 90 days. They included user tracking for abuse and troubleshooting problems, enforcing the one response per computer setting, determining certain respondents being authenticated or tracking the current page of a respondent on a multipage survey and others [39].

Questions were designed after a review of previous survey literature [40–43]. The questionnaire consisted of a set order with a short introduction with explanation of the aim of the study, a section for gathering demographic information anonymously, a main part, and a closing part. [Supplementary file 1 Questionnaire]. The general part was adapted according to Berg et al. [41] and Kazemian et al. [42] and the main part was partly adapted from the survey design by Scheetz et al. on the use of artificial intelligence in ophthalmology, dermatology, radiology and radiation oncology, which was validated through a literature review [43]. The survey also included two open-ended questions: one for suggestions regarding potential AI tools and another for general feedback. Additionally, a back button was available, allowing participants to review and revise their responses. The 20-item questionnaire included Likert scale items, where respondents were asked to rate the extent to which certain statements applied to them. The endpoints of the Likert scales were clearly defined, with “Does not apply at all” and “Applies completely” serving as the extremes for most items. The scale allowed for a nuanced assessment of attitudes and opinions, with intermediate responses “Rather applies”, “neutral” and “Rather does not apply” to capture varying degrees of agreement.

A pilot test of the questionnaire was conducted in advance to ensure effective coverage of the topic, as well as consistency and comprehensibility. The pilot group consisted of five experienced dentists, two of whom use AI in their private practice. The feedback was evaluated, resulting in two minor corrections. This study was conducted in accordance with the Code of Ethics of the World Medical Association [44]. Ethics committee approval was requested from the Ethics Committee of Northwest- and Central Switzerland (EKNZ) in Basel, Switzerland, and was waived since the survey was directed at dentists and no patient data was collected. IP addresses were tracked by SurveyMonkey to prevent multiple entries from the same IP address. No reminders were sent, as the system did not track which participants had completed the questionnaire. Participants’ anonymity was ensured by using only encrypted data, with no personal information being stored. This online survey lasted four weeks and was conducted in autumn 2023.

In terms of statistical analysis, answers to the questionnaire were used via Factor Analysis to create two variables: AI safety (items 1, 2, 3, 7, 8 & 11) and AI utility (items 4, 6, 9 & 10). We performed maximum-likelihood factor analysis on a covariance matrix to identify underlying factors. The Promax rotation method, which allows for correlated factors, was used to enhance interpretability and clarity of the factor structure.

T-tests and ANOVAs were conducted to examine differences in the percentage of correct answers between sex and years of work experience. The correlation of age and opinion on AI was evaluated using Pearson’s product moment correlation coefficient. Additionally, Spearman’s rank correlation coefficient was used to find a correlation between participants who thought AI would positively affect oral health and participants with a positive attitude toward AI. The same measure was used to correlate participants who believe that AI will rather diminish oral health care quality and people with negative attitude towards AI. P-values of < 0.05 were considered statistically significant.

Results

Survey response and data analysis

A total of 129 completed questionnaires were collected and screened for completeness and integrity. After excluding 15 incomplete or incoherent responses, the corrected response rate was ≈ 10.2% (114 out of 1121). The detailed answers of the sex distribution (Q1), years of clinical work experience (Q2), age (Q3), location of practice by canton (Q4), main work constellation (Q5), work percentage (Q6), main field of work (Q7), country of training (Q8), status as a military dentist (Q9) and further results are shown in Supplementary file 2 Results.

The participants’ age ranged from 25 to 88 years, with an average of 46. Of all the participants, 76.3% were male, 20.2% were female, and 1.8% defined themselves as non-binary. Two participants did not provide an answer. A total of 68 (≈ 59.6%) participants had been in practice for more than 16 years, reflecting the extensive experience of dentists in the survey population.

Participants came from various work settings: 52.6% of participants were sole practice owners. 9.6% were working in joint practice with several colleagues as owners. Dentists who worked in joint practices as owner with another dentist and dentists who were employed in joint practices with several dentists were each represented by 8.8%. Figure 1 shows a detailed distribution.

Fig. 1.

Results of “How is the work constellation you’re mainly working in” with the theme work constellation

Dentists in a wide range of specialties were found in the survey population. Fields most widely represented were general dentistry (76.3%), reconstructive dentistry/ prosthodontics (47.4%), endodontology (41.2%) and oral surgery (38.6%) [the term “reconstructive dentistry” is being used mainly in the context of prosthodontics at the University of Basel].

The country of training was Switzerland for almost three quarters of participants with 72.8%. This was followed by 18.4% who studied in Germany and 1.8% who studied in Sweden. The remaining parties made up 6.3%. One participant did not provide the country of training. The canton of Bern had the highest return rate for questionnaires at (27.03% ), followed by the canton of Basel-Land (26.83% ), the canton of Solothurn (16.28%), the canton of Basel-Stadt (15.64%) and the canton of Lucerne (15.56%). A total of 6 cantons had no returns. Four out of these six cantons are French-speaking (Geneva, Vaud, Neuchâtel, and Jura). The canton Ticino, with Italian as first language also had a low response rate (5.26%).

Personal opinion

The statement “My attitude towards AI is fundamentally positive” was agreed upon with ‘rather applies’ by a majority of 54.4% and with ‘applies completely’ by 19.3%. 41.2% of participants rated the statement “I have never given it much thought” with ‘Does not apply at all’, while only 2.6% rated it with ‘applies completely’. Most participants chose ‘Does rather not apply’ concerning the question, whether “AI is rather unappealing” (43.9%) and 54.4% thought “AI is a benefit for our everyday lives” ‘rather applies’. Figure 2 below summarizes the answers graphically; raw data is shown in Supplementary file 2 Results.

Fig. 2.

Results of “Please rate the extent to which the following statements apply” about the opinion on AI (only ‘applies’ and ‘applies completely’ are displayed)

51.8% chose ‘rather applies’ for the statement “Artificial intelligence is already strongly interwoven with our everyday lives”. In comparison, only 0.9% chose ‘applies completely’ for “Artificial intelligence in dentistry is just a trend and has no future”. 35.1% chose ‘does rather not apply’ when confronted with the statement “AI represents competition for humans”. “AI poses a potential threat to humans” was denied by a majority of 32.5%, choosing ‘does rather not apply’. The second most chosen stance was ‘rather applies’ with 27.2%.

“In what time frame could you imagine robots largely replacing the work of doctors/dentists?” was answered with ‘never’ by a majority of 28.1%. Additional answers are illustrated in Fig. 3.

Fig. 3.

Answers to “In what time frame could you imagine robots largely replacing the work of doctors/dentists?”

“Please select the services you have used before” and “Which of these services do you use regularly [>1x/week]” presented several popular services with the implementation of AI technology. The list of services included:

• Communication tools such as translators or large language models (e.g. Deepl or ChatGPT).

• Online shopping (e.g., Zalando, Galaxus or Digitec) [all three of them being popular online shops in Switzerland].

• Music services (e.g., Spotify, YouTube Music, Apple Music).

• Social media or content platforms (e.g., Facebook, Instagram, Tiktok, YouTube, Twitch, etc.)

• Use of biometric data as a key (e.g., mobile phone unlocking using facial recognition or fingerprints).

• Use of AI for image processing (e.g., retouching, filters or image generation).

• AI in dentistry (e.g., caries detection on single tooth x-rays or bitewings).

‘Use of biometric data as a key’ was chosen by a majority of 89.5% for previously used services. In addition, a high percentage of 78.9% of participants stated that they used biometric data as a key at least once a week. Only a single participant said never to have used an AI involving service (Fig. 4).

Fig. 4.

Results of “Please select the services you have used before” and “Which of these services do you use regularly [> 1x/week]” assessing the frequency of usage of AI services

To “How high do you estimate the proportion of dentists in Switzerland who already use AI in caries diagnostics?”, the average response was 13%, with 1–90% estimates.

Future

In the next question “I could imagine innovations in the field of AI in the following areas of dentistry” radiology was on the first place with 87.7% followed by administration (78.9%), implantology (71.9%), and orthodontics (71.1%).

In “Which AI powered support option would you welcome?” participants favoured picture analysis (e.g., differentiation of efflorescences of the oral mucosa) with 90.4%, followed by x-rays (84.2%), assistance in patient management (77.2%) and assistance in research (70.2%).

In “You can choose any AI tool for work, what would be your favourite? (May also be futuristic)” answers ranged from “diagnostic tools” or “orthodontic planning” up to “a tool that detects the kind of pain the patient is having (periodontal, endodontic or articular/ musculoskeletal)” or “a fully automatized secretary, who finishes all administrative work with one mouse click.”

Q19 “Please rate the extent to which the following statements apply” was a final assessment of participant’s opinion on future development of AI in dentistry. “AI will become more important in dentistry” was agreed upon by most with the same number of votes (45.6%) for ‘rather applies’ and ‘applies completely’ each. There was also overall agreement with the statement “AI enriches or will enrich dentistry” with 50% choosing ‘rather applies’ and 36% choosing ‘applies completely’. Finally, “AI worsens or will worsen the quality of modern dentistry” was assessed negatively with 27.2% picking ‘does not apply at all’ and 45.6% picking ‘does rather not apply’.

Statistical analysis

Pearson’s product-moment correlation coefficient (PPMCC) showed a strong correlation between the two created variables AI safety and AI utility (r = 0.51, p < 0.001, Fig. 5). T-tests between the male and female survey population regarding AI safety and AI utility showed no significant differences between sex (p = 0.823 respectively, Figs. 6 and 7).

Fig. 5.

PPMCC of AI safety and AI utility

Fig. 6.

AI safety in correlation to sex

Fig. 7.

AI utility in correlation to sex

The non-binary survey population wasn’t included because the sample size of two was too small to obtain statistically significant results. Analysis through PPMCC showed a weak but significant negative correlation between the age of participants and belief in AI utility (r = -0.19, p = 0.049). AI safety showed no significant correlation between age and factor (p = 0.65).

Spearman’s rank correlation coefficient showed a strong correlation between positive personal attitude towards AI and the conviction that AI enriches or will enrich dentistry in the future (rho = 0.57, p < 0.001). There was also a significant correlation between the negative counterpieces: ‘AI is rather unappealing to me’ and ‘AI worsens or will worsen the quality of modern dentistry’ (rho = 0.43, p < 0.001).

A T-test showed no significant differences in AI utility assessment between military and non-military dentists (p = 0.232). ANOVA also found no significant difference between groups of different years of work experience (p = 0.214). However, there was a significant difference among participants who believed that AI might replace their work in the future and their assessment of AI utility (p = 0.013). Corresponding tests on AI safety didn’t show a significant difference with either military dentists (p = 0.998), work experience (p = 0.559), or the belief that AI might replace work (p = 0.449).

Discussion

Use of and attitude towards AI

AI plays an increasingly central role in the digital world, with private use clearly preceding its professional adoption in dentistry. In Q15, respondents estimated that 13% of Swiss dentists use AI for caries diagnostics. However, 21.9% reported weekly use of AI services in dentistry (Q14), suggesting an underestimation of actual adoption. In Q16, radiology was on the first place with 87.7% followed by administration (78.9%), implantology (71.9%), and orthodontics (71.1%). 63.2% of respondents could imagine innovations in the field of reconstructive dentistry/ prosthodontics. A systematic review from Switzerland by Bernauer et al. (2021) overviewed literature about AI in prosthodontics. Seven studies met inclusion criteria. Six of them reported the training and application of an AI system, one other the function of an intrinsic AI system in a CAD software. The study demonstrated AI developments in prosthodontics and showed its application for automated diagnostics. It concluded that “In the future, AI technologies will likely be used for collecting, processing, and organizing patient-related datasets to provide patient-centered, individualized dental treatment.” [45].

Scheetz et al. reported weekly AI use of 9–20% across medical specialties (ophthalmologists, dermatologists, and radiologists), highlighting variation by field [43].

A high percentage of participants (78.9%) stated that they used biometric data as a key at least once a week. This was followed by 61.4% stating that they use music related services involving AI and 59.6% using social media or content platforms involving AI at least once a week. Overall, the numbers show how omnipresent AI already is in our everyday lives. It matches the answers of Q11 where a majority of 51.8% chose ‘rather applies’ and 14% chose ‘applies completely’ for the statement “Artificial intelligence is already strongly interwoven with our everyday lives”.

Personal opinion towards AI was generally upbeat; a benefit was expected and adverse opinion or indifference was denied in Q10 [statement 1, 2, and 3]. In terms of uncertainty, however, there was a more moderate general sentiment in Q11 [statement 3 and 4].

“AI will become more important in dentistry” was agreed upon by most with the same number of votes (45.6%) for ‘rather applies’ and ‘applies completely’ each in Q19. There was also overall agreement with the statement “AI enriches or will enrich dentistry” with 9.6% choosing ‘neutral’, 50% choosing ‘rather applies’ and 36% choosing ‘applies completely’. These findings are similar to those of other studies: Ivanišević et al. found 71.0% of responding Croatian dentists to believe that AI technologies could enhance patient care [34] and in the study conducted by Hegde et al. targeting Australian dental students and dentists, similar numbers were found with 69% believing that AI would be beneficial in clinical tasks [35].

Scheetz et al. reported similar agreement across medical specialties, with around 70–75% of participants believing AI would improve their field [43].

Regarding dental radiology, recent studies have primarily explored healthcare professionals’ perspectives on AI [30, 34, 38, 41]. Studies by Arora et al. and Bahadir et al. found that patient attitudes toward AI in dental radiology are shaped by education level and context, with a general preference for human oversight but openness to AI as a supportive tool [46, 47]. Comparing these two studies suggests that the way survey questions are framed can significantly influence participants’ perspectives. Arora et al. asked patients whether they preferred diagnosis by AI or by a dentist, with participants favoring dentists. In contrast, Bahadir et al. posed questions regarding the use of AI by dentists versus a diagnosis by dentists alone - resulting in more positive responses towards dentists using AI. This discrepancy highlights how the context in which AI is presented - as a tool for dentists or as replacement of human workforce - can shape patient perceptions. The results further indicate that while respondents generally recognized the potential of AI to enhance dental practice, they also emphasized the importance of human oversight, particularly in diagnostic and treatment-planning applications. This aligns with European efforts to implement ethical guidelines that safeguard patient autonomy and ensure that AI remains a supportive tool rather than a decision-making authority [4, 48].

Referring to the results of this study, it seems intuitive that a positive attitude toward AI strongly correlates with the belief that AI enhances or will enhance dentistry in the future. The strong correlation between the statements ‘AI is rather unappealing to me’ and ‘AI worsens or will worsen the quality of modern dentistry’ further showed how dental professionals base their opinion on the expected benefit of a tool for work. In our study, the t-test between male and female survey population showed no significant difference in p-values in neither AI safety nor AI utility. There is a significant gender gap in computer science. As of 2022, women held only 28% of computing and mathematical jobs in the US [49]. Data from the Swiss Federal Statistical Office from 2022 showed that only 18.5% of university students in IT or Communication studies in Switzerland were women [50].

Analysis through PPMCC showed a significant negative correlation between participants’ age and their belief in AI’s utility (r = -0.19, p = 0.049). This suggests that older dentists tend to be more skeptical about AI’s potential benefits, which may reflect a generational gap in digital familiarity. Another possible reason for rejection of AI technology is a need for more essential knowledge and advanced training in digital medicine [37]. In the study by Pinto Dos Santos et al., which surveyed medical students on their attitude on AI, 71% agreed on the need for AI to be included in medical training [40]. Their questionnaire was later reused by Pauwels et al., which showed that a lecture on AI had a direct positive effect on agreement regarding the different roles of AI in oral radiology. The education also raised overall excitement regarding AI, and concerns about the potential replacement of oral radiologists were decreased [36].

Older generations may face greater digital adaptation challenges, including technophobia, which could explain more cautious views toward AI among experienced dentists [51, 52]. Higher experience might lead to less optimism for new technology and possibly more caution, even though the ANOVA on AI safety assessment and different work experience groups showed no significant difference in participants in this study.

When will AI exceed human performance?

In this paper, 28.1% of participants thought AI could never replace dentists. However, 67.6% agreed that AI powered machines might replace the dentist’s work in 10 to 100 years [Q12]. Our data revealed a significant correlation between participants who believe AI might replace their work in the future and a lower assessment of AI utility (p = 0.013). This indicates that individuals who perceive AI as a potential job threat tend to evaluate its usefulness less enthusiastically, a result that merits further exploration.

In comparison, in the aforementioned study by Pinto Dos Santos et al. only 0,4% of medical students agreed entirely with physicians being replaced in the foreseeable future, 1.1% rather agreed with the statement, and a majority of 81.0% disagreed entirely. “Artificial intelligence will revolutionise medicine in general” was agreed upon entirely by 35.7% and rather agreed upon by a majority of 37.3% [40]. AI researchers Grace et al. put recent development into accurate words: “Advances in artificial intelligence (AI) will transform modern life by reshaping transportation, health, science, finance, and the military”. This subparagraph’s title was taken from the same named paper by them [53]. A more recent study by Grace et al. states that with the current development of AI there is an estimated 50% chance that AI will beat humans in every possible task by 2047, reducing the previously published prediction by 13 years [54]. That expectation coincides with Q12, where 24.6% of participants voted that dentists’ work will be replaced by 20–50 years. Q11 statement 4 “AI poses a potential threat to humans” was evaluated with ‘applies completely’ by 8.8% of participants. Grace et al. found respondents to rather believe in a positive outcome in future AI development, however the likelihood of a bad outcome was assessed - similarly to our findings - to be about 10%, and an “extremely bad (e.g., human extinction)” outcome was assessed to be at 5% [53]. Pauwels at al. found that approximately, one-third of Brazilian dentists and dental students responding to their survey were concerned about AI [36].

Strategic board games like chess, Go, and Shogi were once considered major benchmarks for AI due to their complexity and the need for long-term planning. The rapid progress of AI in these areas, highlighted by the achievements of AlphaZero, which surpassed human-designed programs like Stockfish by teaching itself through reinforcement learning, underscores the increasing capabilities of AI. This shift from AI systems dependent on human input to self-learning entities marks a significant advancement in AI [55–57].

Some studies show that in isolated tasks, AI already has the potential to outperform human professionals [58, 59]: In terms of caries detection, a commercially available AI software achieved perfect sensitivity for both enamel and dentin detection thresholds with a specificity of 0.9 or higher, far exceeding the diagnostic performance of trained dentists in detecting primary caries in bitewing radiographs [58]. Another study similarly found higher diagnostic performance in detection of apical lesions in panoramic dental radiographs by AI compared to radiologists [59].

Law and security

Legal questions always accompany the advent of new technologies; AI is no exception. Legal questions might be one of the leading causes of uncertainties regarding the use of AI in medicine [60].

Modern legislation, such as the European AI Act (AIA), provides a crucial framework for the safe and responsible development of artificial intelligence in medicine and dentistry. It emphasizes transparency, human oversight, and risk mitigation by requiring explainable decision-making processes and professional supervision. These provisions aim to protect patient safety, ensure ethical integration, and support clinicians in benefiting from AI technologies [4–6].

With digitization of paper records, hospital information systems increasingly collect digital data. This data contains health-related data, personal and financial information. Data means money: the official website of the European Union reported a value of 325 billion euro for the EU27 data market in 2019, representing 2.6% of gross domestic product (GDP) [61]. Intelligence platforms Risk Based Security and Flashpoint published a “2021 Year End Report: Data Breach QuickView” that reported 4,145 publicly disclosed security breaches – unauthorized access to data - that exposed over 22 billion records in 2021. The healthcare sector was a viral target in this regard [62].

Healthcare information security has become a big responsibility for healthcare organizations. Reality shows, that while the threat of data breaches is genuine, it’s still an often-overlooked problem [63, 64]. The digital infrastructure of healthcare organizations is frequently the target of hackers. For example, ransomware blocks certain functions or encrypts files until a specific ransom is paid. Data is often accessed from various destinations, simplifying data sharing and joint operations among professionals. It is a strong point of digitalization. Still, it can also make data more vulnerable for outside attacks [65]. An exciting simulation study by Ying et al. showcased the possibility of using AI-supported ethical hacking to uncover weak points in health information systems. Ethical hacking means that hacking is carried out under controlled conditions to detect vulnerabilities that an attacker might use to exploit insufficient safety measures. There were two methods of ethical hacking in the study: One traditional and one optimized by AI. The optimized method outperformed the unoptimized across the board “in terms of average time used, the average success rate of exploit, the number of exploits launched, and the number of successful exploits.” [66]. With this knowledge in mind, it is understandable that about 36% of the participants agreed with the statement “AI is a potential threat to humans” [Q11].

Bias and possible improvement

We believe that our study shed a light on an important topic that had yet to be explored in Switzerland’s academic landscape. As per usual, this process also gave us the chance to reflect on our work and find possible improvements for future research.

The relatively low response rate of 10.2% is a significant limitation of our study and could affect the generalizability of our findings. A low response rate increases the risk of response bias, as non-respondents may have different views from those who chose to participate. Those who responded may have stronger or more polarized opinions about AI in dentistry, which might not reflect the broader population of dental professionals.

Additionally, the demographic profile of the respondents—such as age, gender, and professional background—might not fully represent all Swiss dentists. This discrepancy could further limit the ability to generalize our findings. Clinicians with a strong interest in, or opposition to, AI may have been more likely to complete the survey.

We tried to mitigate these biases by shortening the survey to 5–10 min and ensuring anonymity. However, the digital distribution method, while appropriate for a study on AI, may have also contributed to the low response rate, as it excluded dentists without an online presence. Despite efforts to ensure accessibility, convenience sampling through Google Maps may have missed some dentists, further influencing the response rate. The survey was voluntary and no reminder was sent. A study by Kaplowitz et al. found that response rates can be significantly improved through additional follow-up reminders, suggesting that the absence of multiple reminders in our study may have contributed to the lower participation [67].

However, the response rate of 10.2% observed in our study falls within the range reported in the scientific literature. Studies have shown that the average response rate for e-mail surveys among healthcare professionals is approximately 53%, although substantial variability exists. Factors such as the survey method, the number of reminders sent, and geographic differences can significantly influence response rates [68]. Furthermore, the perceived relevance of the survey topic likely played a role in the response rate. AI in dentistry remains a relatively novel and evolving subject, and despite its growing importance, some dental professionals may not yet recognize its direct relevance to their daily practice. Previous research indicates that survey response rates tend to be higher when participants feel the survey topic is closely related to their professional experience or offers tangible benefits [69]. Some dentists lacked websites or used contact forms, limiting access to their email and possibly excluding digitally less engaged practitioners. For all those reasons, it would be interesting to repeat the survey by postal mail or phone and see how the results differ from the digital survey. Another aspect is the poor response rate from the French (0%) and the Italian-speaking (5.26%) part of Switzerland. Teaching languages in dental schools in Switzerland are either German or French. Therefore, the participants are expected to have a general knowledge of both of these languages. However, for most Ticinesi (people from the canton Ticino), Italian is their mother tongue. German, being the most widely spoken language in Switzerland, was used for the final version of the survey. For future studies, language validation could be expanded to address potential language barriers and improve response rates, particularly from French- and Italian-speaking regions. The existing questionnaire could be translated and distributed to French-speaking clinics to ensure broader participation.

Cultural factors significantly influence attitudes toward AI adoption in dentistry. In our study, Swiss dental professionals operate within a multicultural environment shaped by the country’s diverse linguistic regions. Research indicates that cultural backgrounds impact trust, perceived usefulness, and willingness to adopt AI. Variations in AI acceptance within oral radiology have been linked to educational and cultural factors as shown by Arora et al. [46] Although our study did not explicitly examine these cultural influences, Switzerland’s diverse setting offers a unique opportunity to explore such dynamics. Future research should consider cross-national comparisons to further understand how socio-cultural factors shape AI adoption in dentistry.

The survey population might also show an age and gender bias. 76.3% of participants were male, and the mean age was 46. This distribution likely reflects an incline toward older, male dentists. This may limit generalizability, especially regarding attitudes of underrepresented groups (e.g., women and/or younger dentists). While no significant gender differences were found, the underrepresentation itself remains a methodological limitation. Given the wide age range of respondents (25–88 years), potential outliers may have influenced the observed correlation between age and perceived AI utility, as measured by Pearson’s product-moment correlation coefficient. While all valid responses were included in the analysis, excluding extreme values could have been considered as an alternative to assess the robustness of the result, particularly given PPMCC’s sensitivity to outliers.

The results of our study align with the broader trends observed in recent research concerning dental professionals’ perceptions of AI: In a recent paper from 2024, Dashti et al. did a systematic review on attitude, knowledge and perception of dentists and dental students toward AI. They found that 72.01% of dental students as well as 62.60% of dentists believed in AI’s potential for advancing dentistry, aligning their findings with Fitzek et al., who also found dental students to exhibit slightly more positive attitude toward the integration of AI into their future practices [33]. Dashti et al. concluded their further findings with the goal to promote AI instruction in dental schools and further development of education programs for practitioners [70]. In their follow-up paper, the authors suggest an expanded educational framework with core courses in AI and machine learning, Hands-on training with AI tools and more interdisciplinary collaboration at dental academic institutions [71].

Ivanišević et al. found that major barriers to get and employ AI tools in dentistry included acquisition and maintenance costs (59.0%) and financial constraints (58.0%) [34]. While the financial constraints of Swiss and Croatian dentists may differ, future research could explore this aspect further to understand the factors influencing Swiss dentists’ adoption - or lack thereof - of AI tools in their practice. What motivates investment in AI-driven upgrades or education?

This study focused on opinions and AI usage, but many other important questions remain. What AI tools are currently available on the market? What are their costs, and how well documented is their scientific validity? Additionally, it would be valuable to investigate which AI tools dentists are using, which manufacturers or software providers they prefer, and how satisfied they are with these products. One would need to keep in mind, however, that the development of software has never been advancing faster and creating research that is still up to date due to the delay from the moment of data collection to publication is a challenging task.

The opinion of patients towards AI, as explored by Arora et al. [46] and Bahadir et al. [47] was neither focus of this study, nor did respondents mention it as a point of interest concerning new AI technologies. In times of individualized, patient-centred healthcare, gathering the opinions of patients and comparing them to professional opinions might be of great interest for future research. Mentioned studies focused primarily on radiology in dentistry. Nevertheless, as mentioned, AI can be used in different ways, for example aforementioned drug administration to optimize the selection of drugs for patients and predict harmful interactions [19]. The same mechanisms could allow patient-specific dental care plans, including clear information about what prophylaxis tools to use how often and where in the mouth. It would be interesting to present such approaches as well as established technology like image recognition in dental radiology or oral photography to patients and assess their interest/rejection to it. Possibly also dissecting the gap of trust in between performance of AI itself and the performance of dentists using AI.

Looking ahead, an intriguing hypothesis is that trust in AI will continue to grow, while confidence in dentists who do not incorporate AI may decline.

A study conducted closer to Switzerland in Germany by Kosan et al. additionally showed patients dental radiographs with caries lesion highlighted by an arrow and lesion highlighted by AI-generated coloured overlays. The study found that difference in display to influence patient’s recognition of the lesion. The coloured surface enabled more respondents to recognise the decay [72].

Compared with other surveys conducted in Switzerland (postal or by telephone), our study had a large cohort (in comparison with the telephone survey) and a non-expensive, ecologically better approach compared with postal surveys [73, 74]. The participants received only one e-mail; no reminder was sent. Even if the percentage of returned surveys seems low, 114 entirely conducted surveys still provide a good overview of AI and dentistry in Switzerland.

Conclusions

Our survey showed that 21.9% of respondents already use AI in dentistry at least once a week. A significant negative correlation between age and perceived AI utility (r = − 0.19, p = 0.049) suggests that older dentists tend to be more sceptical. No significant gender differences were observed in perceptions of AI (p = 0.823), but participants who believed that AI might replace jobs in the future were significantly more likely to rate its utility lower.

These findings align with international literature and indicate a cautious optimism within the dental community. They emphasize the growing relevance of AI in clinical dentistry, while also highlighting the need to address generational differences in acceptance through targeted educational strategies.

Although focused on Swiss dental professionals, the results reflect broader, cross-national challenges in AI adoption - particularly within diverse, multilingual healthcare systems like Switzerland’s, where cultural and linguistic contexts may further shape perceptions and implementation.

There are opportunities for further research to deepen our understanding of this topic. For instance, future studies could explore alternative data collection methods, such as conducting surveys via post or phone, to increase response rates and engage a broader sample of dental professionals. Additionally, translating the questionnaire into French and targeting French-speaking dentists could significantly improve response rates and ensure more inclusive participation across Switzerland’s multilingual regions. Furthermore, it would be valuable to investigate patient attitudes towards AI, as this could provide critical insights into how AI adoption in dentistry is viewed by those receiving care. These follow-up studies would help further contextualize the implications of AI in dentistry and guide its future integration into clinical practice.

As AI continues to evolve and integrate into clinical settings, the dental community must not only keep pace with technical advancements but also cultivate thoughtful, responsible engagement. The world doesn’t need smarter machines nearly as much as it needs wiser people using them.

Our findings suggest that targeted education, cross-generational support, and clearer ethical frameworks will be key in unlocking AI’s full potential for dentistry and beyond.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

AI

Artificial Intelligence

AIA

Artificial Intelligence Act

ANOVA

Analysis of Variance

ChatGPT

Chat Generative Pre-trained Transformer

PPMCC

Pearson’s product-moment correlation coefficient

Q

Question

Author contributions

DJ, BIB and KWN conceptualized the study and designed the questionnaire. BJ did the statistical analysis. DJ and BIB wrote the original draft. BM, KWN and TFM contributed to the manuscript and revised the first manuscript. DJ, BIB, BJ and KWN revised the manuscript, BIB was in charge of supervision and administration. All authors have read and agreed to the published version of the manuscript.

Funding

Open access funding provided by University of Basel.

Data availability

Raw data are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

Ethics approval was requested from the Ethics Committee of Northwest- and Central Switzerland (EKNZ) in Basel, Switzerland. In accordance with the EKNZ ethics approval was waived. In agreement with the same ethics committee, data was obtained using a non individualized hyperlink and the aim and purpose of the study were explained in the e-mail containing the link to the questionnaire. By clicking on the link and participating in the study participants gave informed consent. Non consenting recipients could simply ignore the link and e-mail. This study was conducted in accordance with the WMA Declaration of Helsinki – Ethical Principles for Medical Research Involving Human Participants [44].

Consent for publication

Not applicable.

Relevant guidelines and regulations

This study was conducted in accordance with the Code of Ethics of the World Medical Association [44].

Competing interests

The authors declare no competing interests.