How wearable cameras and AI confirm people’s true eating habits

New research suggests that combining wearable cameras, artificial intelligence and biological data could offer the most accurate picture yet of what people eat, addressing long-standing problems in how diets are measured.

In a review published in Nature Food, an international team of scientists argues that no single tool can reliably capture people’s eating habits.

Instead, they say integrating multiple approaches, from digital technologies to biochemical analysis, may provide a far clearer understanding of diet and its links to health and sustainability.

For decades, nutrition research has relied heavily on self-reported methods such as food diaries, questionnaires and recall interviews.

While widely used, these tools often depend on memory, estimation and guesswork, and can be time-consuming for participants.

According to the authors, these weaknesses limit researchers’ ability to accurately connect diet with disease risk, public health outcomes and environmental impacts.

The review brings together evidence from a wide range of fields, including nutrition science, metabolomics, microbiome research, computer vision and sensor technology.

It highlights how recent advances could transform dietary assessment by reducing reliance on self-reporting and increasing objectivity.

One of the study’s co-authors, Dr Thomas Wilson from Aberystwyth University’s Department of Life Sciences, said capturing what people eat remains one of the biggest challenges in nutrition research.

“Traditional methods rely heavily on self-reporting, which we know is imprecise,” he said. “By integrating modern tools such as biological biomarkers and digitally assisted reporting, we can dramatically improve accuracy while reducing the burden on participants.”

Among the technologies discussed are wearable cameras that automatically capture images of meals as they are eaten. Using artificial intelligence and computer vision, these images can be analysed to identify foods and estimate portion sizes, reducing errors caused by forgotten snacks or misjudged quantities.

Smartphone apps can also prompt users in real time, helping to minimise memory-related inaccuracies.

The authors also point to growing interest in biomarkers of food intake, known as BFIs. These involve detecting specific chemicals in urine, blood or stool samples that correspond to particular foods or dietary patterns.

Unlike self-reported data, biomarkers provide objective evidence of what has been consumed, although they cannot yet capture the full complexity of a person’s diet on their own.

Privacy concerns

The review stresses that no single technology offers a complete solution. Wearable cameras may raise privacy concerns, while biomarkers can be costly and are still limited in scope. As a result, the researchers propose an integrated and flexible framework that combines different tools and can be adapted for various settings, from tightly controlled clinical trials to large population-level studies.

According to the authors, improving dietary assessment is essential for the development of precision nutrition, more effective dietary guidelines and evidence-based policies that address both human and planetary health.

Dr Wilson said clearer data on eating habits is increasingly important as societies face rising levels of diet-related disease alongside pressure to adopt more sustainable food systems. “The technologies now emerging give us a real opportunity to build the next generation of dietary assessment,” he said.

The international research was led by scientists from the University of Copenhagen, working with colleagues from Aberystwyth University, the Medical University of Graz, the Institute for Systems Biology, and Wageningen University & Research.