Generative artificial intelligence for counseling of fetal malformations following ultrasound diagnosis

Introduction

The anatomic ultrasound, typically performed between 18 and 22 weeks of gestation, serves as a critical milestone in prenatal care, systematically evaluating fetal anatomy to detect structural anomalies [1], 2]. While most examinations reveal normal findings, the detection of fetal malformations occurs in approximately 2–3% of pregnancies [3]. These diagnoses are often unexpected and profoundly distressing for prospective parents, transforming what was anticipated as a routine, reassuring examination into a moment of shock, uncertainty, and grief. The emotional impact extends beyond the immediate diagnosis, as families must rapidly process complex medical information, understand prognosis, and navigate difficult decisions about pregnancy management and future care. In these pivotal moments, clinicians bear the responsibility not only to deliver accurate medical information but also to provide compassionate support, clear communication, and ongoing guidance throughout the decision-making process.

Generative artificial intelligence (GenAI or GAI) refers to advanced models capable of generating human-like text, images, audio, or code in response to user prompts. Emerging prominently in 2022–2023, models such as GPT-4 (OpenAI), Gemini (Google), Claude (Anthropic), Perplexity, Meta’s LLaMA, and Mistral build on large language models trained on extensive medical, scientific, and general datasets. These tools are increasingly accessible through clinical platforms and web-based interfaces, enabling integration into healthcare workflows with appropriate training and oversight [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14], [15], [16].

For clinicians, GenAI offers novel support in communicating complex and emotionally sensitive medical information. It has demonstrated utility in documentation, patient education, clinical decision support, and medical imaging interpretation [17]. In obstetrics specifically, GenAI enhances patient engagement, comprehension, and continuity of care across diverse clinical scenarios [17], [18], [19]. By translating technical findings, anticipating questions, and reinforcing key messages, it supports more effective and empathetic communication. In fetal counseling, its advantages include 24/7 availability, bias-free synthesis of current literature, and personalized, scalable communication – enabling consistent, high-quality care across varied healthcare settings (Table 1).

Generative AI applications in pregnancy ultrasonography

In perinatal care, before, during, and after an anatomic ultrasound, and especially following the diagnosis of and abnormal ultrasound such as a fetal anomaly, GenAI offers a new layer of support. When applied thoughtfully, it can complement the obstetrician’s role by tailoring communication, improving understanding, and reinforcing empathy during one of the most challenging conversations in obstetrics [15], 16], 20].

Before the ultrasound appointment, Generative AI (GenAI) can help by creating a simple, easy-to-read information sheet for pregnant patients. This sheet can explain what the ultrasound is for, what kinds of things it might show, and what to expect during the visit. Using plain language and clear examples, GenAI can help reduce anxiety and make patients feel more prepared(Table 2).

During ultrasound and prior to formal diagnosis, clinicians can use GenAI to assist in the differential diagnosis of sonographic findings [1], [2], [3]. By entering observed features – e.g., cranial size, echogenicity, or limb abnormalities – models can generate lists of possible anomalies or syndromes, along with literature-based guidance on imaging or evaluation. These tools can enhance diagnostic confidence, support structured reporting, and link findings to relevant coding systems or clinical guidelines. Table 3 illustrates GAI’s ability to enhance diagnostic accuracy by providing instant access to syndrome associations linked to specific fetal anomalies and systematically identifying additional ultrasound features that should be evaluated to confirm or exclude syndromic involvement. In real-time, GenAI can also suggest next steps for targeted imaging, help identify atypical or rare presentations, and flag potential syndromic associations for further consideration. For trainees and less experienced providers, GenAI may serve as an educational support, reinforcing pattern recognition and improving familiarity with anomaly-specific terminology. Used judiciously, these applications can contribute to more consistent, informed, and timely assessment of complex cases.

After the ultrasound, and especially when there is an abnormal ultrasound and a fetal issue has been identified (Table 4), GenAI can help bridge the gap between clinical findings and patient understanding. It can generate individualized summaries of the ultrasound results using plain language, tailored to the specific diagnosis, gestational age, and care plan. These summaries can be reviewed with the patient, printed, or accessed at home – allowing families time to reflect and revisit the information at their own pace. GenAI can also create visual aids, such as labeled diagrams or simplified illustrations of the affected anatomy, which enhance comprehension and make abstract or unfamiliar diagnoses more concrete. In addition, it can generate anticipatory guidance, such as personalized FAQs, likely next steps, and questions to consider during upcoming consultations. For patients facing language or literacy barriers, GenAI can translate content into multiple languages while preserving nuance and clinical accuracy. By supporting clarity, repetition, and emotional preparedness, these tools can reduce confusion, improve shared decision-making, and extend the reach of the clinical team.

Counseling with GAI

Counseling following the detection of a fetal anomaly during the anatomic ultrasound is one of the most sensitive and consequential encounters in obstetrics. Communication must be clear, compassionate, and tailored to the patient’s emotional readiness and level of understanding. Although many institutions have developed protocols for delivering serious diagnoses, the quality and consistency of counseling vary widely. Generative AI can offer valuable support at multiple stages of this process [21], [22], [23], [24].

GenAI can generate layperson summaries that reflect the specific anomaly, gestational age, and management plan – offering families a more accessible and personalized explanation than generic educational material. It can also create visual aids, such as annotated diagrams and simplified anatomical illustrations, to make structural abnormalities more comprehensible. These tools help bridge gaps in health literacy and improve retention of complex information.

Additionally, GenAI can simulate common questions and provide anticipatory guidance in the form of customized FAQ sheets, helping patients prepare for multidisciplinary consultations or genetic counseling. It can translate content into multiple languages while preserving nuance and clinical accuracy, enhancing communication in diverse populations. Used ethically and under clinical oversight, GenAI has the potential to improve patient understanding, reduce decisional conflict, and reinforce empathy during some of the most difficult conversations in obstetric care.

GenAI can be especially valuable when counseling involves specific malformations, such as spina bifida, congenital diaphragmatic hernia, cleft lip and palate, or cardiac anomalies. It is particularly helpful in cases involving rare or unfamiliar anomalies, where clinicians may have limited time or resources to prepare tailored explanations. Each of these diagnoses carries a distinct range of prognostic implications, care pathways, and emotional responses. GenAI can generate condition-specific summaries that are not only medically accurate but also appropriately framed for the patient’s emotional and cognitive readiness. These summaries can be updated in real time as new information emerges – from additional imaging, genetic testing, or multidisciplinary input – ensuring that the content evolves alongside the clinical picture.

For example, in the case of a suspected ventricular septal defect, GenAI can initially provide a general explanation of the condition and potential implications, and later revise the content if follow-up echocardiography confirms isolated vs. syndromic involvement. Importantly, GenAI can help convey both present findings and future steps, such as surveillance plans, likely consultations, and potential postnatal interventions. For families coping with uncertainty, this continuity can reduce confusion and help restore a sense of agency. Whether outlining the schedule for follow-up ultrasounds, providing questions to ask during a pediatric surgical consult, or preparing a summary for a second-opinion visit, GenAI serves as a tool to help families understand the arc of care. When reviewed and personalized by the clinician, this support enhances shared decision-making while preserving the physician’s role as the central guide and interpreter.

Ethical and professional considerations

The deployment of GenAI in obstetric counseling requires careful attention to ethical principles and professional responsibilities [25], 26]. AI-generated content must support, not replace, the physician-patient relationship [27], [28], [29]. The principle of respect for autonomy dictates that patients should remain central in decision-making, with AI tools serving only as enhancements to informed discussions [30], [31], [32]. Clinicians should safeguard against misinformation or errors. GenAI system do not – and cannot – decide what is true and what is false. Sometimes, they just confabulate, a phenomenon known as hallucinations, especially in rare or borderline conditions [33], 34]. As of June 2025 the newest and most powerful technologies – so-called reasoning systems from companies like OpenAI, Google and the Chinese start-up DeepSeek – are generating more errors, not fewer.

Consequently, any content created by GenAI must be reviewed and approved by a qualified clinician. Similarly, beneficence requires that AI outputs be helpful and supportive, not overwhelming or distressing. Tone, length, and readability must be optimized to avoid information overload. Professional accountability remains paramount. GenAI tools should never be presented as authoritative by themselves. The clinician must remain the interpreter of medical findings and the guide for decision-making. Finally, considerations of equity must inform implementation. Access to GenAI-enhanced counseling should not be limited by geography, language, or socioeconomic status.

How to get the most of GenAI

Generative AI is not autonomous. It does not generate useful or clinically appropriate content on its own. Instead, its effectiveness depends entirely on the expertise and direction provided by the clinician. In the context of fetal counseling, GenAI functions as a responsive tool – its value is determined by the quality of the input it receives. To harness its full potential, clinicians must understand how to structure prompts, define clinical objectives, and critically review the output. This requires not only medical knowledge but also familiarity with the capabilities and limitations of GenAI systems [17], 35], 36].

A prompt is the set of instructions, context, or questions a user provides to the AI to elicit a specific output [37]. Much like framing a clinical consultation or asking a colleague for input, the clarity, completeness, and tone of the prompt directly influence the relevance and quality of the AI’s response. In clinical settings, effective prompts must be purposefully designed to reflect patient needs, medical facts, and counseling goals.

Prompt engineering refers to the design of specific instructions to guide GenAI outputs. In ultrasound counseling, this technique is critical to ensure that generated content is accurate, empathic, and aligned with clinical goals [38]. Poorly designed prompts can result in misinformation, inappropriate tone, or irrelevant content.

There is a clear evolving science of how to create effective prompts. Effective prompts must include a multitude of clinical context, patient demographics, diagnosis, and desired tone (e.g., neutral, supportive, hopeful). For example, a prompt to generate a lay explanation of tetralogy of Fallot might include the specific findings, planned interventions, and the emotional state of the family. The output should be informative yet comforting, avoiding alarmist or overly technical language.

Prompt engineering also enables localization – the tailoring of content to regional policies [15], cultural norms, and legal frameworks. This is especially relevant when discussing options like termination or perinatal palliative care, which vary significantly across jurisdictions. Clinical teams must be trained in constructing effective prompts and reviewing outputs for fidelity to the medical facts and the values of the patient.

How GAI can help delivering difficult news: the need for sensitivity and support

Receiving a diagnosis of a fetal malformation is among the most distressing experiences an expectant parent can face [39]. What was anticipated as a routine ultrasound can abruptly become a moment of shock, fear, and grief. In these emotionally charged situations, the clinician’s sensitivity is not ancillary – it is essential. Patients confronted with such news are often overwhelmed by the complexity of the medical information and the weight of decisions ahead. In these moments, what patients most need is not only clinical clarity but human compassion. How the information is conveyed – tone, pacing, word choice – can shape a family’s understanding, emotional response, and long-term trust in the healthcare system. Cultivating clinical empathy in these encounters is therefore both a moral and professional imperative [40], 41].

Empathy is a foundational element of clinical excellence. Patients consistently rate empathetic physicians as more competent, trustworthy, and effective, even when clinical outcomes remain unchanged [42], 43]. In emotionally charged settings – such as the disclosure of a fetal anomaly – empathy plays a vital role in building trust, reducing anxiety, and facilitating shared decision-making. It conveys not only understanding, but presence and commitment, and strongly influences how patients remember and respond to the counseling experience.

Patients often express dissatisfaction not with the medical care itself, but with what was traditionally called a clinician’s “bedside manner” – a concept now better understood as empathy. One frequent critique of technology in medicine is its perceived inability to replicate empathy. However, when guided by skilled prompts and clinician oversight, generative artificial intelligence (GenAI) can enhance – not diminish – perceived empathy by reinforcing the human elements of care. Empathic language, patient-centered phrasing and anticipatory validation of emotions can be integrated into GenAI-generated content, helping families feel seen and supported [44], [45], [46].

Emerging evidence suggests that GenAI may outperform humans in perceived empathy under certain conditions. In experimental studies, third-party evaluators and patients have rated AI-generated responses to emotionally sensitive scenarios as more empathic than those of clinicians, particularly in written formats [47], [48], [49]. These findings, while preliminary, highlight the potential of GenAI to reinforce core elements of compassionate care – especially when used as a supplement to, not a substitute for, clinician communication. Language models can be tuned to avoid clinical detachment, reflect emotional resonance, and tailor phrasing to patient distress, thereby strengthening the therapeutic alliance.

Moreover, GenAI can reinforce these messages after hours or between visits, offering consistent empathic support at moments of uncertainty or distress. When families review materials at home or share them with others, the embedded tone and language continue to reflect the clinician’s commitment. Used responsibly, GenAI becomes a tool for extending – not replacing – the empathic voice of the care team.

Families often leave counseling sessions feeling overwhelmed and uncertain about where to turn next. One of the notable strengths of GenAI is its ability to curate and deliver relevant support tools to patients. GenAI can compile condition-specific educational material from reputable organizations, ensuring that the information is both current and comprehensible [50].

In cases of poor prognosis, families may require introductions to perinatal palliative care or hospice planning. GenAI can deliver such content with sensitivity and clarity, emphasizing that support continues even in the absence of curative options. In more treatable conditions, GenAI can connect patients with early intervention services, surgical centers, and advocacy organizations, offering guidance tailored to the specific diagnosis and stage of care.

Families also frequently seek peer support. GenAI can identify and suggest moderated forums, support groups, and local or national organizations that align with the family’s needs and values. By personalizing these resources, the technology can meaningfully extend the reach and effectiveness of the care team.

Case examples and sample prompts

1. Fetal diaphragmatic hernia. For example, a well-constructed prompt might read: “Generate a compassionate, easy-to-understand explanation of a diagnosis of congenital diaphragmatic hernia at 22 weeks’ gestation. Begin by acknowledging that this is difficult news to hear. Use reassuring and supportive language to explain the condition, next steps, and how the care team is committed to walking alongside the family throughout the process.” Such prompts yield responses that open with emotional validation – e.g., “We understand that hearing this diagnosis can be overwhelming…” – and continue with clear, gentle explanations of the diagnosis and its implications. This combination of clarity and compassion reflects what patients value most in high-quality, human communication.

2. Anencephaly. The diagnosis is devastating, and the prognosis is uniformly fatal. A GenAI tool could generate a gentle, comprehensible summary for the family explaining the anomaly, expected clinical course, and local options for palliative care or pregnancy termination. The content could be aligned with the family’s language preference, religious considerations, and access to care. For example, a brief prompt might read: “Write a compassionate, easy-to-understand explanation for parents whose fetus has been diagnosed with anencephaly at 22 weeks. Acknowledge the emotional difficulty, explain the diagnosis and prognosis clearly, and describe options for care, including comfort-focused pathways. Use gentle, supportive language throughout.” This type of prompt can help ensure that the generated content balances medical clarity with emotional sensitivity, matching the gravity of the situation while offering reassurance and support.

3. Isolated mild ventriculomegaly. The prognosis is uncertain and depends on the progression and presence of other findings. Here, GenAI could produce a summary emphasizing the need for follow-up imaging, potential neurologic outcomes, and support resources. The family could receive a printable set of questions for their next consultation, increasing their agency and engagement. A sample prompt might read: “Write a supportive, easy-to-understand summary for expectant parents following a prenatal diagnosis of mild isolated ventriculomegaly. Explain what this means, the importance of follow-up ultrasounds or MRIs, and possible outcomes. Include 3–5 helpful questions the parents might ask their provider.”

4. Cleft lip and suspected cleft palate. The anomaly is correctable postnatally, but families often need reassurance and planning support. GenAI could provide diagrams, links to surgical centers of excellence, and a stepwise timeline from prenatal planning to pediatric surgery consultation, easing the transition between specialties. A brief prompt could be: “Generate a reassuring explanation for parents whose fetus has a prenatal diagnosis of cleft lip with possible cleft palate. Emphasize the condition’s treatability, outline expected care steps from birth to surgical correction, and include a calm and supportive tone.”

As shown in Table 5, effective GenAI implementation requires carefully crafted prompts that specify the clinical scenario, gestational age, desired communication tone, and specific counseling objectives to generate appropriate, empathetic responses tailored to each unique fetal diagnosis and family situation.

Future directions and research needs

Counseling following the prenatal diagnosis of a fetal malformation is one of the most delicate and consequential conversations in obstetric care. These moments are filled with uncertainty, vulnerability, and the urgent need for accurate, compassionate guidance. Generative artificial intelligence, when applied thoughtfully and under clinical supervision, can significantly augment this process. Rather than replacing the clinician’s role, GenAI can serve as a powerful extension of the healthcare team’s voice and empathy.

To realize the potential of GenAI in perinatal counseling, several priorities must be addressed. First, rigorous evaluation of GenAI-generated content is essential. Clinical validation studies should assess accuracy, readability, and emotional impact. Second, systems should include feedback mechanisms so that patients and providers can report inaccuracies or suggest improvements.

Integration with electronic health records (EHRs) will be important for scalability. This could allow automatic generation of summaries or resources based on imaging findings. Collaboration with professional organizations could help standardize templates, train clinicians in AI use, and ensure equitable access across health systems.

Research should also examine whether GenAI reduces decisional conflict, improves satisfaction, or supports better outcomes for families facing complex diagnoses. Ethical frameworks must continue to evolve alongside the technology.

By helping clinicians generate accurate, personalized, and culturally sensitive explanations, GenAI can support both the informational and emotional needs of patients and their families. It can reinforce key messages outside the clinical setting, translate complex findings into accessible language, and link patients to reliable educational resources and psychosocial support. Importantly, GenAI has the potential to democratize access to high-quality counseling materials, bridging gaps in health literacy, language, and location.

Ethical implementation is paramount. When embedded into a framework of clinical oversight, patient feedback, and transparent communication, GenAI becomes a tool for deepening trust rather than undermining it.

As perinatal medicine continues to evolve, GenAI stands not as a replacement for human presence but as a meaningful adjunct in the art and science of prenatal counseling. With careful design and ethical stewardship, generative AI can amplify the clinician’s ability to inform, comfort, and guide families through some of the most challenging decisions they may ever face, for transforming the counseling of fetal malformations identified by ultrasound. When designed and used ethically, with attention to empathy, cultural relevance, and clinical accuracy, these tools can enhance communication, support patient understanding, and improve access to high-quality resources. Rather than replacing the human connection, GenAI can amplify the capacity of clinicians to inform, comfort, and guide families through difficult and often life-altering conversations.