MicrotiaGPT: Appraisal of Microtia Care Recommendations From an Artificially Intelligent Chatbot Authors: Oluwatobiloba Ayo‐Ajibola, Simon Park, Beth Osterbauer, Krystina Kallemeyn, Gabriel Gomez Published: 2026 ABSTRACT ABSTRACTObjectiveEvaluate the accuracy, comprehensiveness, and similarity to provider response of ChatGPT‐4o in providing patient education regarding microtia and aural atresia management.MethodsTen standardized inquiries were created across three domains (General Information, Hearing/Anatomy, Treatment Decision‐Making) and entered into ChatGPT‐4o. Responses were evaluated by an international multidisciplinary cohort of 23 experts, including pediatric otolaryngologists and audiologists. Using 5‐point Likert scales, graders assessed factual accuracy, comprehensiveness, similarity to physician responses, and overall quality. Readability was analyzed using Flesch–Kincaid metrics.ResultsChatGPT‐generated responses demonstrated high accuracy (mean = 4.03 [SD = 0.66]), comprehensiveness (mean = 4.18 [SD = 0.54]), and similarity to the provider's response (4.02 [0.70]). Audiologists assigned significantly lower ratings than pediatric otolaryngologists regarding accuracy (mean = 3.76 vs. 4.13; p < 0.001) and comprehensiveness (3.94 vs. 4.30; p = 0.046). Qualitative analysis included concerns regarding advanced terminology, omission of specific risks, and a lack of empathetic framing regarding parental guilt. Flesch–Kincaid Reading Grade Level often exceeded the 5th‐grade average level for patient education materials (mean = 9.64 [SD = 1.21]).ConclusionChatGPT‐4o provides reliable introductory information but lacks the technical nuance and multidisciplinary coordination essential for complex microtia care. While a useful adjunct for early information‐seeking, ChatGPT‐4o lacks the multidisciplinary framing and depth required for ideal counseling. Introduction Microtia is a congenital underdevelopment of the auricle, ranging from minor structural deficiencies to complete absence of the ear. Concomitant ear canal atresia and conductive hearing loss are found in 90% of patients [1]. The combination of a facial deformity and hearing loss impacts the development and psychosocial well‐being of patients with microtia. Children with microtia report increased prevalence of depression, social withdrawal, and teasing [2]. Additionally, the parents of children with microtia commonly express feelings of guilt and concerns over psychosocial implications [3, 4]. While parents continue to seek medical providers with microtia expertise, families frequently use the internet as a source of information. As of 2017, more than two‐thirds of American adults reported using the internet as the first resource in their search for health information [5]. Use of the internet for medical information has grown beyond search engines. In a survey of online health information users, 20% reported using ChatGPT, an artificially intelligent chatbot as a source. Of these, 80% reported the tool to be just as useful as or more useful than their doctor. Given these sentiments, it is likely the public's use of artificial intelligence, AI, for their health education needs will only grow [6]. Research across medical specialties has started to emerge on AI chatbot capabilities. Within otolaryngology, ChatGPT has been found to provide comprehensive and accurate responses that align with those of otolaryngologists treating salivary gland disorders and obstructive sleep apnea [7, 8]. However, there are criticisms regarding the appropriateness and specificity of outputs from this platform [9]. This study aims to provide a comprehensive assessment of ChatGPT's efficacy in microtia education via a novel approach by incorporating an international, multidisciplinary cohort of microtia‐focused clinicians and a clinically relevant grading system. It is hypothesized that, at this point in its training, ChatGPT‐4o will display moderate comprehensiveness, accuracy, and similarity to physician responses, but will likely lack specificity when qualitatively analyzed. Representative preoperative and postoperative outcomes of autologous rib cartilage graft reconstruction, one of the two primary surgical approaches discussed in patient counseling, are shown in Figure 1. Rib graft reconstruction: (a) Preoperative view of the microtia ear before rib cartilage reconstruction. (b) Postoperative result following rib cartilage graft reconstruction. Materials and Methods This cross‐sectional cohort study was deemed exempt from Institutional Review Board review at the University of Southern California based on the lack of interaction with protected patient populations or use of health‐related information. Inquiry Creation Ten inquiries about the care of microtia and atresia were developed based on the clinical experience of a board‐certified pediatric otolaryngologist (GG) (Table S1). These 10 were categorized into three groups: “general information,” “hearing and anatomy,” and “treatment decision‐making.” General information questions were based on initial information seeking including queries such as “what is microtia?”, “what is atresia?” and “what causes microtia?” Questions related to hearing and anatomy aimed to simulate information‐seeking about physical details and hearing such as: “Does my child have an ear canal?”, “what kind of hearing loss will my child have?”, and “Can the ear canal be opened?” Finally, the treatment decision‐making category aimed to mimic patient inquiries regarding management options including “Does my child need scans or imaging?”, “What are the options for improving my child's hearing?”, “How can my child's ear be reconstructed?” and “Which is better, Medpor or rib reconstruction?” Survey Creation Each inquiry was entered into a blank ChatGPT inquiry session in a browser window lacking history‐taking capability. This was done to obtain de novo ChatGPT's responses and avoid skewed answers. The first generated response from ChatGPT‐4o, queried on May 15, 2025, was recorded for each question and compiled in Table S1. A RedCAP survey was created with each ChatGPT response displayed on an individual page as well as three criteria for evaluation. Factual accuracy measured the perceived number of factual errors in the response, from 1 (totally inaccurate) to 5 (totally accurate). Comprehensiveness determined how well each answer fulfilled the request, graded on a 5‐point Likert scale from 1 (very noncomprehensive) to 5 (very comprehensive). The similarity to provider response evaluated the level of congruence each provider grader perceived with their prior experience answering such questions, using a similar 5‐point Likert scale from 1 (very dissimilar) to 5 (very similar). Finally, the overall quality of the response was graded on a 5‐point scale where 5 was very high quality and 1 was very low quality. Survey Distribution and Expert Evaluation This convenience sampling survey was distributed electronically via an established international microtia specialists' WhatsApp group, which serves as a centralized communication platform for clinicians with a focus in microtia and atresia management. Readability Measurement Readability was evaluated using the Flesch Reading Ease (FRE) and Flesch–Kincaid Reading Grade Level (FKRGL) algorithms present on Microsoft Word for Microsoft 365 (Microsoft Corporation, Redmond, WA). The FKRGL is correlated to an equivalent grade level, whereas FRE scores are graded along a scale in which lower scores indicate a higher education level. Although ChatGPT is capable of providing answers of a desired readability level based on a prompt, no such prompts were utilized in this study to mimic patient interactions, as readability specification is not likely in these scenarios. Statistical Analysis Data analysis was performed using Microsoft Excel (version 16.71, Microsoft Software, Richmond, Washington) and Stata (Stata Statistical Software: Release 18. College Station, TX: StataCorp LLC). Descriptive statistics characterized the data, and chi‐squared and ANOVA tests were employed to compare scores and readability between categories of inquiries and subspecialty graders. If ANOVA was significant, post hoc testing was further conducted to analyze intergroup variability. Significance was set at p < 0.05. Results In total, 23 respondents completed the assessment, including 12 pediatric otolaryngologists (52.17%), 5 audiologists (21.74%), 2 facial plastic and reconstructive surgeons (8.70%), 2 otologists (8.70%), and 2 respondents did not specify (8.70%) (Table 1). On a 5‐point Likert scale, ChatGPT‐generated responses demonstrated moderately high accuracy (mean = 4.03 [SD = 0.66]), comprehensiveness (mean = 4.18 [SD = 0.54]), and similarity to the provider's response (4.02 [0.70]). Clinicians exhibited moderate agreement on accuracy (κ = 0.46), fair agreement on comprehensiveness (κ = 0.35), and treatment decision‐making (κ = 0.39) domains. TABLE 1Overall evaluator and content.Characteristic (n = 23 graders) n (%) or mean (SD)SpecialtyPediatric Otolaryngology12 (52.17%)Facial plastic and reconstructive surgery2 (8.70%)Otologist2 (8.70%)Audiologist5 (21.74%)Not specified2 (8.70%)Overall content quality scores (5 pt likert scale)Accuracy4.03 (0.66)Comprehensiveness4.18 (0.54)Similarity to physician response4.02 (0.70)Test of interrater reliability (Fleiss Kappa)Accuracy0.46Comprehensiveness0.35Treatment decision‐making0.39 Overall evaluator and content. Across the subtypes of microtia‐related inquiries, the quality of ChatGPT responses differed significantly in accuracy (p = 0.045), comprehensiveness (p = 0.026), and similarity to physician responses (p = 0.007) (Table 2). Clinicians assessed the responses to queries about general information on microtia as the most accurate (4.22 [0.16]), most comprehensive (4.24 [0.18]), and most similar to the physician's response (4.07 [0.07]). Responses regarding treatment decision‐making were found to have the lowest accuracy (3.95 [0.13]). Responses about hearing and anatomy had the least comprehensiveness (4.07 [0.16]) and least similarity to a physician response (3.96 [0.08]). Readability metrics placed the responses at a high‐school‐to‐early‐college reading level: FRE scores ranged from 49.2 to 58.5, FKRGL scores ranged from 8.7 to 10.9, and the Gunning Fog Index scores ranged from 12 to 14.1. TABLE 2Question‐level content quality metrics and readability analysis by content domain.QuestionMean accuracy (mean, SD)Mean comprehensiveness (mean, SD)Mean similarity to physician response (mean, SD)Flesch–Kincaid reading easeFlesch–Kincaid reading levelGunning fog indexGeneral informationMicrotia4.35 (0.78)4.35 (0.65)4.13 (0.81)53.71013.3Atresia4.04 (0.82)4.09 (0.85)4.00 (0.90)43.411.814.6Cause of microtia4.26 (0.75)4.43 (0.59)4.09 (1.04)50.510.214.3 Category mean 4.22 (0.16)4.24 (0.18)4.07 (0.07)49.210.914.1Hearing and anatomyType of hearing loss4.00 (0.90)4.13 (0.76)4.00 (0.85)579.912.9Canal opening3.96 (0.88)3.91 (0.79)4.00 (0.80)559.212.4Canal presence3.96 (0.88)4.00 (0.80)3.87 (1.01)56.910.513.8 Category mean 3.97 (0.03)4.07 (0.16)3.96 (0.08)56.39.913Treatment decision‐makingImaging4.09 (1.00)4.22 (0.80)4.17 (1.07)64.37.310.5MEDPor vs. rib recon4.00 (0.85)4.22 (0.67)4.17 (0.83)65.18.712.1Options for hearing3.78 (1.00)4.13 (0.76)3.65 (1.30)558.911.9Reconstruction options3.91 (0.85)4.22 (0.74)4.09 (0.95)49.69.913.6 Category mean 3.95 (0.13)4.22 (0.04)4.02 (0.25)58.58.712 p value 0.045 0.026 0.007 0.162 0.075 0.079 Note: Bolded values signify statistical significance of p < 0.05. Question‐level content quality metrics and readability analysis by content domain. Note: Bolded values signify statistical significance of p < 0.05. The two most represented respondent types, pediatric otolaryngologist and audiologist, had significantly different ratings. Audiologists provided lower ratings than pediatric otolaryngologists across several domains such as overall accuracy (4.13 [0.74] vs. 3.76 [0.32]; p < 0.001) and comprehensiveness (4.30 [0.55] vs. 3.94 [0.27]; p = 0.046; Tables 3 and 4). In domain‐specific analysis audiologists rated general information answers as less comprehensive (4.47 [0.63] vs. 3.87 [0.18]; p = 0.004). For hearing and anatomy‐related content, audiologists also rated the content to be less accurate (4.11 [0.76] vs. 3.67 [0.41]; p < 0.001) and less similar to how the evaluating audiologist would respond (4.06 [0.63] vs. 3.80 [0.61]; p = 0.045). Finally audiologists rated treatment decision making outputs as less comprehensive (4.00 [0.25] vs. 4.25 [0.63]; p = 0.027), compared to participating pediatric otolaryngologists. TABLE 3Question‐level content quality metrics by specialty.CategorySpecialty groupMean accuracy (mean, SD)Mean comprehensiveness (mean, SD)Mean similarity to physician response (mean, SD)OverallPediatric Oto4.13 (0.74)4.30 (0.55)4.18 (0.72)FPRS/otology4.10 (0.93)4.05 (0.80)3.88 (1.09)Audiology3.76 (0.32)3.94 (0.27)3.76 (0.46) p value0.6100.4440.543General informationPediatric Oto4.28 (0.74)4.47 (0.63)4.28 (0.62)FPRS/otology4.33 (0.72)4.17 (0.69)4.00 (0.98)Audiology4.00 (0.41)3.87 (0.18)3.73 (0.64) p value0.6980.1570.346Hearing and anatomyPediatric Oto4.11 (0.76)4.19 (0.48)4.06 (0.63)FPRS/otology3.92 (1.00)3.83 (0.84)3.75 (0.96)Audiology3.67 (0.41)3.93 (0.60)3.80 (0.61) p value0.5370.4890.663Treatment decision‐makingPediatric Oto4.04 (0.87)4.25 (0.63)4.19 (0.92)FPRS/otology4.06 (1.12)4.12 (0.92)3.88 (1.36)Audiology3.65 (0.34)4.00 (0.25)3.75 (0.40) p value0.6570.7540.640 Question‐level content quality metrics by specialty. TABLE 4Comparison of question‐level content quality metrics between pediatric otolaryngology and audiologists.CategorySpecialty groupMean accuracy (mean, SD)Mean comprehensiveness (mean, SD)Mean similarity to physician response (mean, SD)OverallPediatric Oto4.13 (0.74)4.30 (0.55)4.18 (0.72)Audiology3.76 (0.32)3.94 (0.27)3.76 (0.46) p value < 0.001 0.046 0.183General informationPediatric Oto4.28 (0.74)4.47 (0.63)4.28 (0.62)Audiology4.00 (0.41)3.87 (0.18)3.73 (0.64) p value0.339 0.004 0.149Hearing and anatomyPediatric Oto4.11 (0.76)4.19 (0.48)4.06 (0.63)Audiology3.67 (0.41)3.93 (0.60)3.80 (0.61) p value < 0.001 0.417 0.045 Treatment decision‐makingPediatric Oto4.04 (0.87)4.25 (0.63)4.19 (0.92)Audiology3.65 (0.34)4.00 (0.25)3.75 (0.40) p value0.211 0.027 0.095 Note: Bolded values signify statistical significance of p < 0.05. Comparison of question‐level content quality metrics between pediatric otolaryngology and audiologists. Note: Bolded values signify statistical significance of p < 0.05. When provided an opportunity to provide subjective rationale for lower ratings, respondents most often cited issues related to terminology, omission of risks, excessive information, and failure to mention the most updated hearing devices (Table 5).Representative preoperative and postoperative outcomes of Medpor (porous polyethylene) implant reconstruction, the alternative reconstructive approach evaluated in clinician responses, are shown in Figure 2. TABLE 5Qualitative themes identified in clinician feedback with representative comments.ThemeRepresentative commentsNumber of clinicians highlighting issue (n)Terminology and clinical accuracy “Use porous polyethylene instead of Medpor.” “BAHA, Ponto, Adhear are models, not brands.” “Microtia is never painful.” “Clarify microtia vs. atresia.” 18Excessive or misfocused detail “Too much information; overwhelming for families.” “Comprehensive but about the wrong things.” “I would provide less detail.” “Way too verbose and confusing.” 15Omission of risks, options, or updates “Add risk of facial‐nerve injury.” “Include 3‐D printed or cadaveric cartilage options.” “Mention that canal surgery often doesn't improve hearing.” “Update implantable device list (Osia, Sentio, Bonebridge).” 16Lack of patient‐centered tone and communication “Avoid implying parental blame for microtia.” “Address guilt before discussing causes.” “Too textbook—needs empathetic phrasing.” “Simplify language for non‐physicians.” 11Coordination and sequencing of care “Discuss timing of CT and reconstruction.” “Mention collaboration with audiology and plastics.” “Integrate treatment steps into a timeline.” “Add rationale for multidisciplinary approach.” 9 Qualitative themes identified in clinician feedback with representative comments. “Use porous polyethylene instead of Medpor.” “BAHA, Ponto, Adhear are models, not brands.” “Microtia is never painful.” “Clarify microtia vs. atresia.” “Too much information; overwhelming for families.” “Comprehensive but about the wrong things.” “I would provide less detail.” “Way too verbose and confusing.” “Add risk of facial‐nerve injury.” “Include 3‐D printed or cadaveric cartilage options.” “Mention that canal surgery often doesn't improve hearing.” “Update implantable device list (Osia, Sentio, Bonebridge).” “Avoid implying parental blame for microtia.” “Address guilt before discussing causes.” “Too textbook—needs empathetic phrasing.” “Simplify language for non‐physicians.” “Discuss timing of CT and reconstruction.” “Mention collaboration with audiology and plastics.” “Integrate treatment steps into a timeline.” “Add rationale for multidisciplinary approach.” Medpor implant reconstruction: (a) Preoperative view of the microtia ear before Medpor implant reconstruction. (b) Postoperative result following Medpor implant reconstruction. Discussion This study provides a multidisciplinary evaluation of ChatGPT‐4o as a tool for patient education in microtia and aural atresia. The model generated responses that were consistently rated as moderately high in accuracy, comprehensiveness, and similarity to clinician counseling. However, these aggregate strengths are balanced by important domain‐specific limitations that are relevant to real‐world patient care. Performance was strongest for general information queries and declined in areas requiring nuanced clinical judgment, multidisciplinary coordination, and individualized counseling. A notable finding of this study was the difference in ratings between pediatric otolaryngologists and audiologists. Audiologists assigned significantly lower scores across multiple domains, particularly for content related to hearing and anatomy. These differences likely reflect variation in domain‐specific expertise and clinical priorities. Audiologists' focus on hearing rehabilitation and device optimization may be more attuned to omissions in device‐specific counseling. In contrast, pediatric otolaryngologists may evaluate responses more broadly within the context of reconstructive surgery and function. Additionally, publicly available online medical information—on which large language models are trained—may disproportionately reflect surgeon‐authored content, potentially underrepresenting audiology‐specific perspectives. The contrast in ratings underscores the complementary roles of the surgeon and audiologist in delivering the highest‐quality care to the individual patient. Prior studies evaluating AI‐generated responses in microtia have lacked important viewpoints. Both Dadi et al. and Liao et al. evaluated the performance of large language models and found moderate levels of comprehensiveness and accuracy [10, 11]. These studies used combinations of plastic surgeons and/or patients for evaluation; neither included otolaryngologists nor audiologists. A unique strength of the present study is the incorporation of an international, multidisciplinary panel of experts, allowing for a more comprehensive and clinically representative assessment of content quality. This approach captures variability in subspecialty viewpoints that would not be apparent in the prior single‐discipline (plastic surgery) evaluations. The present study provides a more nuanced appraisal of AI performance in the wholistic and team‐based care approach to microtia. This study further supports ChatGPT as an accessible tool for early patient education. Prior work has demonstrated low caregiver satisfaction with information provided during initial clinical encounters, with reported rates as low as 35% [12]. Given the relative rarity of microtia and limited access to specialized providers, ChatGPT may serve as a useful adjunct to supplement early information needs. However, chatbot‐generated information remains inherently limited compared with in‐person, multidisciplinary care. Clinical decision‐making in microtia is highly individualized and depends on factors not captured by current AI interactions, including patient‐specific anatomy, developmental status, family preferences, surgeon‐specific reconstructive philosophy (e.g., autologous versus alloplastic approaches), and site‐specific resource availability [13, 14]. In‐person encounters also provide critical elements beyond information delivery, including empathy, reassurance, and responsiveness to emotional and social cues. These factors remain essential in counseling families navigating complex, longitudinal conditions such as microtia. Conclusion ChatGPT provided reliable introductory information on microtia and aural atresia for common patient and caregiver inquiries. While these tools may serve as useful adjuncts for early information‐seeking, they lack the multidisciplinary framing and depth required for optimal counseling. In their current form, AI‐generated microtia responses require clinician oversight and further refinement to support comprehensive patient education. Funding The authors have nothing to report. Disclosure No artificial intelligence or large language model tools were used in the writing, editing, or preparation of this manuscript. Conflicts of Interest The authors declare no conflicts of interest. Supporting information Table S1: ChatGPT responses to simulated microtia reconstruction inquiries.