Introduction

Challenges in Reproductive Medicine

Despite four decades of progress since the first IVF baby, ART faces significant hurdles. Infertility affects over 186 million individuals globally, with demand for fertility services rising due to delayed childbearing and disruptions from the COVID-19 pandemic, which delayed treatments during lockdowns.1 High treatment costs, often exceeding $10,000 per IVF cycle, lengthy procedures, and variable success rates—ranging from 20–40% per cycle and plateaued over the past decade—limit access and efficacy.2 These challenges highlight the urgent need for innovative solutions to enhance clinical outcomes and affordability in reproductive medicine.

The Promise of Artificial Intelligence

AI holds transformative potential for ART by leveraging large datasets to improve efficacy and reduce subjectivity.3 AI applications in embryo selection, gamete assessment, and personalized treatment protocols aim to optimize outcomes and streamline care, though ethical concerns like algorithmic bias must be addressed.4 This manuscript reviews AI’s role in ART, evaluating its clinical and economic impacts, ethical considerations, and future directions.

Figure 1
Figure 1.Overview of AI applications across the ART pipeline.

Artificial intelligence supports each stage of the assisted reproductive technology (ART) process. In diagnostics, AI enables data-driven intake assessments and risk stratification. During ovarian stimulation, it optimizes gonadotropin dosing based on patient-specific variables. In gamete assessment, AI analyzes oocyte and sperm morphology and motility with enhanced precision. For embryo selection, AI integrates morphokinetic, genetic, and clinical data to rank embryos by viability or ploidy. Finally, in patient counseling, predictive models assist in estimating treatment outcomes and tailoring protocols, improving personalization and shared decision-making.

Literature Search Methodology

A systematic review of peer-reviewed articles published between 2019 and 2025 was conducted using databases including PubMed, Scopus, and Web of Science. Search terms such as “artificial intelligence,” “assisted reproductive technologies,” “in vitro fertilization,” and related concepts were combined using Boolean operators (e.g., AND, OR) to identify relevant studies. Articles were sourced from high-impact journals in reproductive medicine. Studies were included based on their focus on AI applications in ART, specifically embryo selection, gamete assessment, personalized protocols, genetic testing, outcome prediction, and workflow optimization. Exclusion criteria encompassed non-peer-reviewed sources, non-English studies, and irrelevant topics. A total of 152 studies were screened, with 42 included for analysis. Priority was given to studies evaluating key AI tools (DeepEmbryo, icONE, iDAScore, ERICA) and their performance metrics, ensuring a focus on quantitative outcomes.

Results

AI Applications in ART

AI has significantly advanced ART, improving accuracy, clinical outcomes, and efficiency across multiple processes. Below, we detail key applications, supported by quantitative data, while noting validation scopes and limitations.

Embryo Selection

AI-driven embryo selection tools surpass traditional morphological assessments by offering objective, data-driven evaluations.

Table 1 summarizes leading tools’ performance.

Table 1
AI Tool Description Accuracy/Outcome Reference
DeepEmbryo Analyzes static embryo images using Convolutional Neural Network with transfer learning 75% clinical pregnancy prediction accuracy (single-center) 5
icONE Integrates genetic and clinical data for embryo viability 92% implantation accuracy, 77.3% clinical pregnancy rate (multicenter) 6
iDAScore Grades embryos via time-lapse videos Matches manual outcomes, 60% euploid prioritization (multicenter) 7
ERICA Ranks embryos by ploidy status 51% biochemical pregnancy rate for “optimal” embryos (limited validation) 8,9

These tools reduce subjectivity, with icONE achieving a 77.3% clinical pregnancy rate compared to 50% in non-AI groups.6 iDAScore matches manual assessments while reducing evaluation time by 30%, enhancing laboratory efficiency.7 However, most outcomes reflect surrogate endpoints, with live birth rates underreported.

Personalized Treatment Protocols

AI optimizes drug selection and dosing using patient-specific variables, improving ovarian stimulation and reducing treatment cycles.10,11 AI-optimized protocols reduced follicle-stimulating hormone (FSH) usage by up to 20%, potentially lowering medication costs.12

Gamete Assessment

AI improves sperm and oocyte quality analysis, enhancing fertilization success. Image analysis algorithms assess morphological features, increasing sperm motility assessment accuracy by 15% compared to human evaluations, thus reducing variability.13,14

Outcome Prediction and Workflow Optimization

AI predicts IVF success rates, supporting patient counseling and treatment planning.15 Full-cycle management systems integrating machine learning and IoT reduced laboratory processing time by 35%, cut costs by 25%, and increased pregnancy rates by 12%.16

Table 2 summarizes these applications.

Table 2
Application Technology Impact Reference
Personalized Protocols Machine learning Optimized drug dosing, 20% FSH reduction 10,11
Gamete Assessment AI-based image analysis 15% improved sperm/oocyte selection accuracy 13,14
Workflow Optimization Machine learning, Internet of Things (IoT) 35% faster processing, 25% cost reduction 16

Despite these advancements, small sample sizes and single-center studies limit generalizability for some applications.

Table 3.Comparison of AI Tools in Assisted Reproductive Technologies (ART)
AI Tool Function Reported Accuracy/Outcome Validation Scope Reported Outcome Type Estimated Cost/Infrastructure Regulatory Status Clinical Adoption
DeepEmbryo Embryo image analysis (CNNs) 75% pregnancy prediction accuracy Single-center, retrospective Clinical pregnancy (surrogate) Moderate (software only) Pre-clinical; not FDA/CE listed Experimental/research use
icONE Embryo viability prediction using clinical + genetic data 92% implantation accuracy, 77.3% pregnancy rate Multicenter, industry-sponsored Implantation and pregnancy rate (clinical) High (genomic + clinical integration) Limited CE use; research setting Early clinical use in select centers
iDAScore Embryo grading via time-lapse analysis Matches manual grading, 60% euploid prioritization Multicenter pre-clinical validation Euploidy prioritization, manual agreement (surrogate) Moderate (time-lapse system required) CE-marked, clinical pilot use Adopted in European fertility centers
ERICA Embryo ranking by ploidy prediction 51% pregnancy rate for ‘optimal’ embryos Limited clinical validation Pregnancy rate and ploidy rank (surrogate) Low to moderate (data analysis layer) No formal clearance reported Pilot usage in AI-focused clinics

side-by-side comparison table of leading AI tools in ART, summarizing their function, performance, validation scope, cost, regulatory status, and clinical adoption

Discussion

Clinical and Economic Impacts

AI enhances ART by improving objectivity, personalization, and efficiency. Tools like icONE and ERICA reduce subjective judgment, achieving clinical pregnancy rates of 77.3% and 51% (biochemical), respectively, compared to 50% in non-AI groups.6,8 Personalized protocols optimize outcomes, reducing the need for multiple IVF cycles and alleviating emotional and financial burdens.2 AI-driven workflow systems, such as sperm bank management, cut laboratory costs by 25% (e.g., ~$2,000 per cycle) and improve efficiency by 35%.16 However, high initial investments in software and training may limit access in smaller clinics.17 Longitudinal cost-benefit analyses are needed to confirm AI’s economic viability across diverse healthcare systems.18

Ethical and Regulatory Considerations

AI integration in ART raises several ethical challenges, particularly around data privacy, algorithmic transparency, and equity. Tools like DeepEmbryo and icONE process sensitive genetic and clinical data, necessitating compliance with data protection frameworks such as the EU’s General Data Protection Regulation (GDPR), which mandates informed consent and algorithmic accountability.5,6,19 In contrast, the U.S. Health Insurance Portability and Accountability Act (HIPAA) permits more flexible data sharing under certain conditions, leading to variability in AI deployment across jurisdictions.20

A major ethical concern is algorithmic bias resulting from non-representative training datasets. Many AI models are developed using data predominantly from Western or homogeneous populations, which may reduce accuracy and reliability when applied to diverse patient groups. This underrepresentation risks exacerbating existing disparities in access and outcomes, as AI systems may underperform in populations with different genetic backgrounds, socioeconomic profiles, or cultural contexts. Ensuring the inclusion of diverse, global datasets is essential to avoid perpetuating inequities in fertility care.7,21

Regulatory oversight is also evolving. Tools like iDAScore have received CE mark certification in Europe, while others remain in pre-clinical or pilot phases. Regulatory frameworks such as the FDA’s “Software as a Medical Device” (SaMD) guidance are essential for ensuring safety, especially as AI algorithms become adaptive and dynamic over time.22,23 However, current approval processes are often ill-suited to continuously learning systems, highlighting the need for harmonized global standards that balance innovation with patient protection.24

Robust ethical frameworks, emphasizing transparency, fairness, and inclusivity, are critical for the responsible deployment of AI in ART. Collaboration between clinicians, data scientists, ethicists, and regulators will be essential to address these complex issues and foster trust in AI-assisted reproductive care.

Patient Perspectives

Patient trust is critical for AI adoption in ART. Tools like ERICA and icONE improve outcomes, but fears of depersonalization may hinder acceptance.6,8

The study of Cromack et al.provides relevant data on the trust of IVF patients in AI. This study surveyed 200 patients undergoing IVF or frozen embryo transfers, assessing their demographics, technological affinity, and perceptions of AI in fertility care. The results indicated that while 93% of respondents were familiar with AI and 55% supported its use in medicine, only 46% trusted AI-informed reproductive care. Additionally, patients showed a preference for physician-based recommendations over AI in treatment-related decisions, although a notable proportion favored AI for gamete and embryo selection compared to gonadotropin dosing or stimulation length.25

This study provides a comprehensive overview of patient trust in AI within the context of fertility treatments, highlighting both the potential and the reservations patients have regarding AI applications in this field.

Technology Integration

AI’s potential is amplified by integration with genomics, wearables, and robotics. icONE’s genomic-clinical data fusion achieves 92% implantation accuracy, improving euploid embryo selection.3,6 Wearables enable real-time hormonal monitoring, personalizing stimulation protocols, while AI-driven robotics in sperm banks boost efficiency by 35%.10,16

Longitudinal Studies and Real-World Impact

AI’s long-term impact, particularly on live birth rates (LBR), remains understudied. DeepEmbryo’s 75% clinical pregnancy prediction and iDAScore’s 60% euploid prioritization lack diverse, real-world validation.5,7 ERICA’s 51% biochemical pregnancy rate requires LBR confirmation.8 iDAScore’s 46.5% clinical pregnancy rate slightly underperforms morphology-based selection (48.2%), while ERICA’s 0.79 positive predictive value for euploidy surpasses embryologists.26,27 Longitudinal studies should assess patient satisfaction, equity, and standardized metrics to ensure sustainable benefits.18,28

Study Limitations and Publication Bias

Many AI tools (e.g., DeepEmbryo, ERICA) are validated in single-center studies, limiting generalizability across diverse populations.22,29 Most studies report surrogate endpoints (e.g., clinical pregnancy) rather than LBR, the definitive ART success measure.18 Publication bias may introduce optimism bias in reported outcomes, particularly in industry-sponsored studies like icONE, may inflate efficacy; trial preregistration could mitigate this.30 Large-scale, multicenter trials are critical for robust, equitable AI application in ART.31

Conclusion

Artificial intelligence has catalyzed a transformative shift in assisted reproductive technologies, optimizing embryo selection, gamete assessment, personalized protocols, and workflow efficiency. Advanced tools achieve high clinical pregnancy rates and implantation accuracy, minimizing subjectivity and enhancing outcomes. Despite substantial initial infrastructure costs, AI-driven systems reduce expenses and improve efficiency, alleviating the financial burden of IVF cycles. However, challenges persist, including limited validation scopes, reliance on surrogate endpoints, and ethical concerns such as data privacy and algorithmic bias. Regulatory disparities and publication bias further hinder equitable adoption. To fully harness AI’s potential, large-scale, multicenter trials must prioritize live birth rates, patient satisfaction, and inclusive datasets. Integration with genomics, wearables, and robotics offers prospects for precision fertility care, though interoperability and affordability remain critical barriers. As global infertility rates rise, AI’s continued evolution, driven by rigorous validation, ethical frameworks, and harmonized regulations, will be essential for delivering accessible, equitable, and effective reproductive healthcare.

Future Directions in AI for Reproductive Medicine

  • Conduct large-scale, multicenter validation studies to assess AI tool performance across diverse populations and clinical settings.

  • Standardize outcome reporting by prioritizing hard clinical endpoints such as live birth rate, cumulative live birth per cycle, and long-term neonatal outcomes.

  • Develop robust ethical frameworks to address data privacy, algorithmic transparency, and equity in access to AI technologies.

  • Advance regulatory harmonization across jurisdictions (e.g., FDA, CE mark) to streamline approval and ensure patient safety.

  • Incorporate patient-reported outcomes (PROs) to evaluate the psychological, emotional, and experiential aspects of AI-guided fertility care.

  • Ensure inclusivity in algorithm training datasets to minimize bias and improve generalizability for underrepresented populations.

  • Explore AI integration with emerging technologies such as genomics, wearables, and robotic systems for next-generation fertility solutions.

  • Promote interdisciplinary collaboration between clinicians, data scientists, ethicists, and policymakers for sustainable AI deployment.

Conflict of interest

None to declared

Financial support

No financial support was received for this project