AI Helps Deter Wine Fraud

In the world of wine retailing, where the origin of a grape can massively impact the value of the finished bottle, the temptation to pass off one provenance of grape in favour of another is overwhelming.

Nuances of taste often elude even the most seasoned palates, which is why scientists from the University of Geneva (UNIGE) and the Institute of Vine and Wine Science at the University of Bordeaux ¹ have ventured into uncharted territory: decoding the chemical signature of wines with the help of artificial intelligence (AI).

Wine, a complex amalgamation of thousands of molecules, owes its distinct flavour profile to factors such as soil composition, grape variety, and the techniques employed by winemakers. With climate changes, evolving consumer preferences, and an uptick in counterfeiting, the ability to pinpoint a wine's origin has become imperative.

Traditional methods, like gas chromatography, have been employed to separate and analyse the components of a wine. However, the sheer number of molecules in wine creates an intricate chromatogram, making it challenging to discern meaningful patterns. Enter AI, specifically machine learning, a tool that excels at identifying recurring patterns in complex datasets.

The researchers, led by Alexandre Pouget from UNIGE, utilised chromatograms from 80 red wines spanning twelve vintages and originating from seven Bordeaux estates. Instead of focusing on specific peaks in the chromatograms, they employed machine learning to consider the complete data, eliminating unnecessary variables and reducing dimensionality.

The outcome was a revelation: when the researchers plotted the new coordinates on a graph, distinct ‘clouds’ of points emerged, each representing the chemical signature of wines from different estates. Astonishingly, this method achieved a 100% accuracy in identifying the geographical origin of a wine.

Contrary to the expectation that specific molecules would define a wine's identity, the researchers found that a wine's chemical fingerprint was defined by a broad spectrum of compounds. This insight challenges the conventional belief that a few key components determine a wine's uniqueness.

Beyond the scientific breakthrough, the implications are profound. This research not only sheds light on the intricate components contributing to a wine's identity but also presents opportunities for practical applications. The newfound ability to identify the origin of wines with precision can aid in decision-making within the wine industry, ensuring the preservation of a region's unique terroir.

Additionally, the development of such tools could prove instrumental in combating the escalating issue of wine counterfeiting. As the demand for fine wines continues to grow, the risk of fraudulent practices has become a significant concern. The application of AI in authenticating the origin of wines could serve as a powerful deterrent to counterfeiters.

1 - https://www.nature.com/articles/s42004-023-01051-9