The researchers in multiple European countries aimed to evaluate how a six-week personalized nutrition program, delivered via an AI-powered app, would impact the gut microbiome.
They also looked at changes in food intake, body measurements, and biochemical markers, and how these changes relate to shifts in gut microbial composition.
The study, which was funded by a large EU funding program (PROTEIN under Horizon 2020) aimed at supporting research, innovation, and science projects across Europe, was published in the journal Nutrients.
AI nutrition apps
AI has become increasingly important in nutrition, with the technology being used to develop personalized diets. AI tools like chatbots, virtual dietitians, and health apps have shown benefits in managing weight, diabetes, and cancer-related dietary needs.
AI-driven systems - using machine learning, wearables, and mobile apps - can now track diet and provide tailored recommendations. However, according to the researchers, their effectiveness in accounting for individual traits like lifestyle, health data, or the gut microbiome, needs more research.
According to the researchers, current programs tend to overlook biological variability and focus too heavily on genetics, despite limited evidence linking genes to dietary behavior.
“Most of the personalized nutrition concepts have focused on genetic variants, but without adding evidence on how genetics influences dietary behavior, as noted in the Food4Me European study, the largest study of personalized nutrition to date,” they wrote.
“Given the advances in the field of omics sciences, the exploration of other key biological factors such as the gut microbiome will offer unprecedented opportunities and open novel avenues in the field of personalized nutrition.”
The microbiome can explain much of the variability in how people respond to food, and as diet shapes the gut microbiome, it, in turn, affects metabolism and food responses. As a result, personalized diets that target the gut microbiome have been linked to better cardiometabolic health and cognitive function.
Previous research has also shown promising results when microbiome data guides personalized nutrition, including improved glucose and triglyceride responses and better management of irritable bowel syndrome.
Study details
The researchers recruited 29 healthy individuals in Thessaloniki, Greece, and guided them through a six-week intervention using the AI-powered PROTEIN app, which provided personalized meal and physical activity plans.
At the start, a nutritionist helped participants set diet and exercise goals tailored to their needs. The app delivered daily plans based on individual profiles, while a dietician provided ongoing support.
Participants completed a validated food frequency questionnaire (FFQ) to record their usual diet and the International Physical Activity Questionnaire (IPAQ) to assess activity levels. The researchers measured participants’ height, weight, body fat, muscle mass, metabolic rate, and body circumferences.
The researchers collected fasting blood samples at the start and the end to analyze biochemical markers and calculate insulin resistance (HOMA-IR), beta cell function (HOMA-B), and the atheromatic index. They also analyzed participants’ fecal samples before and after the intervention in order to analyze gut microbiota.
After six weeks, participants significantly reduced their intake of carbohydrates (by 18.1%), proteins (13.2%), and total energy (12.7%). They also cut back on alcohol (39%), sweets (33%), and fast food (14%). Waist circumference decreased by 1.5% (an average reduction of 1.2 cm), though weight, BMI, and muscle mass showed no significant changes.
Participants also showed significant shifts in gut microbiota. The intervention increased the number of beneficial bacteria, particularly Eubacterium coprostanoligenes group and Oscillibacter, which have been linked to cholesterol reduction and cardiovascular health. At the same time, it reduced potentially harmful genera like Eubacterium ruminantium group and Gastranaerophilales, which have ties to inflammation, obesity, and disease risk.
The intervention also influenced SCFA-producing microbes. It increased beneficial genera like Faecalibacterium and Subdoligranulum, known for anti-inflammatory effects and metabolic health benefits.
Initial data showed that about 70% of participants consumed more calories than EFSA recommendations, while physical activity (PA) levels varied: 27.6% showed low activity, 51.7% moderate, and 20.7% high. Sitting time averaged 6.6 hours daily, with 44.8% classified as sedentary (≥8 hours/day).
However, physical activity levels improved modestly throughout the intervention period, with the proportion of individuals classified as having low activity falling from 27.6% to 13.8%. Notably, five out of eight participants in the low-activity group transitioned to a higher activity level.
“AI-driven programs may facilitate research into food-based therapies by identifying dietary interventions that support a healthy microbiome and advance the use of the food as a medicinal approach,” the researchers concluded.
They noted that future studies should integrate multi-omics data to refine personalized nutrition further, and researchers should also explore long-term effects on chronic conditions like obesity and diabetes and consider environmental sustainability in dietary recommendations.
Source: Nutrients 2025, 17(7), 1260; https://doi.org/10.3390/nu17071260; “The Influence of an AI-Driven Personalized Nutrition Program on the Human Gut Microbiome and Its Health Implications.” Authors: Rouskas, K. et al.