New method traces roasted coffee origins using chromatography and AI

Roasted coffee may finally have a forensic trail. A team in Italy and the United States published a workflow that pairs comprehensive two-dimensional gas chromatography with computer vision to identify a coffee’s geographic origin after roasting has wiped out many of the usual visual and chemical cues.

The paper appeared in Journal of Chromatography A on May 10, 2026, after an early online release on Feb. 21. Its author list included Giorgio Felizzato, Eloisa Bagnulo, Giulia Tapparo, Giorgia Botta, Qingping Tao, Stephen E. Reichenbach, Chiara Cordero, Luciano Navarini, Erica Liberto and Andrea Caratti, with affiliations spanning the University of Turin, GC Image LLC in Lincoln, Nebraska, the University of Nebraska–Lincoln and illycaffè S.p.A. in Trieste. The project was backed by Horizon Europe under Grant Agreement No. 101135764, placing the work inside the European Union’s main research and innovation programme for 2021 to 2027, which carries an indicative budget of EUR 93.5 billion.

The science matters because roasted coffee is hard to authenticate once heat, blending and distribution have done their work. The paper says a coffee’s volatile profile is shaped by botanical origin, climate, soil, post-harvest treatments and roasting parameters, and that those profiles can include hundreds of compounds. The workflow starts with untargeted fingerprinting, then builds composite class images from multiple chromatograms, followed by targeted peak extraction and multivariate analysis to pick out discriminant compounds. In plain English, it tries to read the coffee’s smell signature after the roast has made the bean much harder to read.

That has direct implications for premium single-origin labeling, fraud detection and consumer confidence. If a roaster is selling a lot as Ethiopia, Colombia or Guatemala, a post-roast authentication tool could help back up the claim when the paper trail is thin and the price premium depends on trust. It could also strengthen supply-chain verification for importers and exporters that need a defensible way to settle disputes over origin.

The University of Turin group was already pushing the idea in 2025. A conference abstract from the same team said it tested 32 commercial roasted and ground coffees from Brazil, Colombia, Ethiopia, Guatemala and India, using computer vision based on GCxGC-MS/FID data. That gives the method a concrete test bed beyond the theory.

The field is not standing still, either. A separate 2025 study from China used an e-nose to trace roasted coffees back to origin, showing that post-roast authentication is becoming a race between different lab tools. This new GC×GC workflow aims to make that race more visual, more chemical and more useful when the roast has already covered the bean’s tracks.