AI-driven personalized nutrition shows promise in improving gut health

By Dr. Priyom Bose, Ph.D.Reviewed by Benedette Cuffari, M.Sc.Apr 10 2025

A six-week pilot study reveals that tailored diets powered by artificial intelligence may enhance gut microbiome diversity and reduce diet-related health risks, though more research is needed.

Study: The Influence of an AI-Driven Personalized Nutrition Program on the Human Gut Microbiome and Its Health Implications. Image Credit: Davor Gabor / Shutterstock.com

Artificial intelligence (AI)-based personalized nutrition programs have the potential to positively influence the gut microbiome in humans. However, additional research is needed to establish the use of gut microbiome-informed tactics for personalized nutrition.

A recent Nutrients study examine the impact of AI-based personalized nutrition programs on the gut microbiome of healthy individuals.

The health benefits of personalized nutrition

A healthy diet and lifestyle is crucial for reducing the risk of non-communicable diseases such as diabetes, cancer, obesity, and cardiovascular diseases. Despite these guidelines, the rates of diet-related diseases continue to rise, which could be attributed to significant variability in how each individual responds to food. Thus, there remains an urgent need for new personalized strategies as an alternative to the ineffective ‘one-diet-fits-all’ approach.

Over the past several years, researchers have become increasingly interested in the potential of personalized nutrition plans to alleviate health conditions like cardiometabolic diseases and promote healthy aging.

AI technology in nutrition

Food scientists and nutrition experts have recently implemented AI technologies to promote sustainable, environmentally friendly, and personalized diets. For example, AI-driven chatbots have been developed to create optimal diet plans for weight loss and manage diabetes, whereas an evidence-based AI virtual dietitian has recently been generated to address diet-related questions for cancer patients.

For personalized nutrition, robust machine learning models can support digital health systems, wearable sensors, and mobile applications, which are now being monitored to assess the effectiveness of generated dietary recommendations customized for an individual’s needs and characteristics.

However, modern personalized nutrition programs appear to underestimate the importance of biological factors that influence the variability of an individual’s responses to food with regard to their health.

About the study

The researchers of the current study assessed the effects of a six-week intervention of an AI-based personalized nutrition program on the gut microbiota composition of healthy individuals. Diet-driven changes in macronutrient levels, anthropometric and biochemical traits, as well as other gut microbial modifications were also evaluated.

A pilot study of twenty-nine healthy participants recruited from the Center for Research and Technology (CERTH) in Greece was conducted. Within the framework of the PROTEIN project, selected candidates were instructed on how to use a digital smartphone health application that provides guidance for maintaining a healthy, nutritionally sound, and active lifestyle.

The PROTEIN mobile app provided daily and weekly meal recommendations based on a novel AI personalized nutritional advisor. This AI-based application considers the user’s dietary preferences, health conditions, and physical characteristics to suggest personalized appropriate dietary plans.

At baseline (pre-PROTEIN), the nutritionist and study participants set dietary and physical activity (PA) goals, which could be attained through an active lifestyle and adherence to a Mediterranean diet designed to the individual’s specific needs. Personalized nutrition and PA plans were automatically generated by the AI advisor and delivered to participants through the PROTEIN app on smartphones.

After this period (post-PROTEIN), the nutritionist evaluated the participants’ progress at the follow-up visit. 

Study findings

The mean age of the study cohort was 35 years, all of whom lived above the poverty line. Most study participants were married and non-smokers. Out of 29 individuals, 20 exceeded the recommended daily energy intake.

Genomic sequencing of V3-V4 regions of 16S ribosomal ribonucleic acid (rRNA) was performed on 58 samples collected from 29 individuals. A total of three phyla, 19 classes, 44 orders, 82 families, and 231 genera were identified.

Firmicutes and Bacteroidota were dominant gut microbial microbiomes identified at baseline and the six-week follow-up visit. At both time points, Prevotella, Bacteroides, and Faecalibacterium were frequently identified. However, higher gut microbiota diversity and abundance were observed at the post-PROTEIN timepoint as compared to pre-PROTEIN baseline levels.

Rhodospirillales was the most upregulated amplicon sequence variants (ASVs), which were ranked by significance, followed by Eubacterium coprostanoligenes group and Ruminococcus genera. The functional potential of the observed taxonomic alterations was assessed through metagenomic analyses, which identified 12 pathways with nominal significance, most of which were associated with microbial metabolic processes and purine degradation.

The post-PROTEIN timepoint was associated with a significant reduction in carbohydrate, protein, and total energy intake. Mean decreases of 39%, 33%, and 14% in the intake of alcohol/beverages, sweets, and fast food, respectively, were also observed at the end of the intervention. Notably, adherence to the Mediterranean diet did not change between time points.

No significant changes in anthropometric measurements were observed except a small but significant reduction in the mean waist circumference. PA levels were consistently variable both pre- and post-PROTEIN among the study cohort.

Changes in sweet intake positively correlated with body weight, fat, hip circumference, and hemoglobin measurements. A robust positive association between fat intake and the abundance of Oscillospiraceae was observed.

A strong positive association between urea and Lachnospiraceae was observed, whereas a negative correlation between cholesterol levels and Oscillibacter was reported.

Conclusions

AI-supported personalized dietary interventions have the potential to promote overall health by facilitating healthy proliferation of the gut microbiome. In the current study, these alterations in the gut microbial ecosystem led to reduced constipation, bloating, and inflammatory bowel syndrome symptoms, while also supporting immune function.

To validate these findings and provide a holistic evaluation of the impact of AI-based personalized nutrition approaches, future studies with longer follow-up periods and larger sample sizes are needed.