Researchers from Frontiers in Public Health tested popular AI chatbots—including ChatGPT, Claude, Gemini, and DeepSeek—by asking them identical questions about child obesity risk, but varying details about the child's background. The study revealed that these AI systems systematically rated risk differently based on demographic information rather than health facts alone.
The children assessed were varied across race, ethnicity, family income level, and whether they lived in urban or rural areas. This real-world experiment exposed how AI tools used by parents for health guidance carry hidden assumptions built into their training data and design.
Income bias is the most consistent problem across all systems
Every single AI model rated low-income children as higher obesity risk, even when given identical health information about diet, exercise, and sleep. This held true regardless of which AI system was used or where it was developed. Income was the strongest predictor of bias in the study.
Race and ethnicity biases appear across nearly all health domains
Most AI systems—whether made in the West or China—consistently assigned higher obesity risk to Black and Hispanic/Latino children compared to white children. This bias showed up in how the AI interpreted diet quality, physical activity levels, and sleep patterns. The bias was widespread but not universal across every single assessment.
Urban-rural assumptions differ by where the AI was built
Western-made systems like ChatGPT tended to flag rural children as higher risk, while Chinese-made systems flagged urban children as higher risk. There is no medical reason for this flip; it reflects cultural assumptions embedded in each system's training data. This inconsistency shows that geography bias is learned from regional patterns, not rooted in actual health science.
Before you ask an AI chatbot about your child's weight
- Recognise that identical health information may produce biased risk assessments based on your child's race, income level, or neighbourhood type.
- Use AI chatbot advice as general information only, not as a substitute for assessment by your child's doctor.
- If you use AI tools, cross-check any concerning findings with a healthcare professional who knows your child's individual circumstances.