Google Research has demonstrated that an artificial intelligence model can estimate how far along a pregnancy is about as accurately as expert sonographers, even when scans are performed by healthcare workers without specialized ultrasound training. The findings could help expand access to prenatal imaging in communities where trained specialists and medical equipment remain difficult to access.

Announced on October 21, 2026, the prospective, multicenter study evaluated Google‘s AI-assisted ultrasound technology in clinical settings in Chicago, Illinois, and Nairobi, Kenya. The research showed that the model maintained comparable accuracy across the two locations, including in Nairobi, where researchers did not fine-tune the model using local patient data.

The results, published in JAMA Network Open, build on Google’s efforts to make ultrasound-based pregnancy assessments more accessible through portable equipment and AI-powered analysis.

How Google’s AI-assisted ultrasound works

Traditional pregnancy ultrasound examinations require trained sonographers to capture specific views of the fetus and take precise measurements. This expertise takes years to develop, creating a significant barrier in areas where healthcare facilities lack specialized staff.

Google’s approach uses a technique called a blind sweep ultrasound. Instead of requiring an operator to locate specific fetal structures and capture precise cross-sectional images, the operator moves a portable ultrasound probe across the abdomen in a predefined pattern.

The resulting scan videos are analyzed by an AI model that estimates gestational age, or how far along the pregnancy is. The system can also help identify inadequate scans that need to be repeated.

According to Google, healthcare workers with no previous ultrasound experience can learn the scanning procedure in approximately eight hours. This could allow midwives and frontline community health workers to perform basic pregnancy assessments without first completing the extensive training required to become professional sonographers.

The research system also processes the scans directly on a mobile device, allowing it to work without an internet connection. Combined with handheld ultrasound equipment, that capability could make the technology particularly useful in remote clinics and other healthcare settings with limited infrastructure.

Study finds accuracy comparable to expert sonographers

For the new study, researchers trained the AI model using ultrasound data from 1,968 participants in North Carolina and Zambia. They then adapted it using data from 120 participants in Chicago before testing it prospectively in two clinical environments.

The evaluation included 192 participants at Northwestern Medicine in Chicago and 193 at Jacaranda Health’s community clinic network in Nairobi. The primary analysis covered 385 participants between 16 and 36 weeks of pregnancy.

The researchers compared the AI model’s gestational age estimates against the standard clinical approach, which uses measurements taken by expert sonographers.

Across both clinical settings, the AI model achieved a mean absolute error of 4.18 days, compared with 4.48 days for the expert-sonographer approach. A lower mean absolute error indicates that estimates were closer to the reference measurements on average.

The results were also consistent across the two locations. The model recorded a mean absolute error of 4.10 days in Chicago and 4.27 days in Nairobi.

Importantly, the researchers did not use data from Nairobi to fine-tune the model. This independent evaluation offers evidence that the system can generalize beyond the population and clinical environment used to adapt it.

Google also reported promising exploratory results in pregnancies involving fetuses with estimated weights above the 90th percentile, where the AI model outperformed traditional biometry formulas. However, these findings are an additional research observation rather than proof that the system performs better across all pregnancy assessments.

Why the findings matter for maternal healthcare

Access to prenatal ultrasound remains uneven around the world. Google cites estimates indicating that roughly two-thirds of people globally lack access to diagnostic medical imaging, including ultrasound.

That shortage matters because gestational age helps healthcare professionals plan prenatal care, assess pregnancy progress, and make decisions about treatment and delivery. When ultrasound services are unavailable, clinicians may have fewer reliable tools for establishing how far along a pregnancy has progressed.

An AI-assisted system that works with inexpensive, portable equipment could help bring some of these assessments closer to patients. Instead of requiring every clinic to have a dedicated sonographer, healthcare facilities could potentially train existing frontline workers to acquire scans and use AI to interpret them.

The Chicago and Nairobi results are particularly relevant because the locations differ in their clinical workflows, patient populations, and access to routine prenatal imaging. The model’s performance across both environments suggests that the approach is not limited to a single hospital or healthcare system.

Still, comparable accuracy in estimating gestational age does not mean AI can replace sonographers or perform every function of a comprehensive ultrasound examination. The study evaluates a specific task, and its findings should not be interpreted as evidence that the system can independently diagnose all pregnancy complications.

Real-world deployment still needs safeguards

Google’s research also highlighted practical challenges that emerge when AI-powered medical technology moves from controlled experiments into everyday clinical use.

For example, researchers found that swappable batteries could be more practical than plug-in docking stations when devices need to be used for consecutive patient appointments. Such operational details can make a meaningful difference in clinics where charging infrastructure and electricity supplies are unreliable.

The company also emphasized the importance of building trust with healthcare workers and patients. Its researchers identified confidence estimates and safeguards that reject poor-quality scans as useful ways to communicate uncertainty and prevent unreliable inputs from producing misleading results.

These safeguards are important because a system that performs well in a study must also demonstrate that it can be used safely and consistently in different clinical environments.

Google has presented the findings as a step toward broader access to ultrasound-based maternal care, rather than an announcement of a widely available commercial medical product. Further validation, careful implementation, and appropriate clinical oversight will be important before this approach can deliver its potential benefits at scale.

For now, the study provides encouraging evidence that AI-assisted ultrasound could help frontline healthcare workers gather useful pregnancy information in places where specialist expertise is scarce. If the approach proves reliable in broader deployments, it could help more expectant mothers access an important part of prenatal care without having to travel to a facility with a dedicated ultrasound specialist.