Words On Wine - Autumn 2025

Words by Robert Black

My AI sommelier described the wine as “having a lovely bouquet, well-structured tannins, and a distinct aroma of… burning rubber. Apparently, it was aged in a Tesla.”

This was the first response from my Samsung Gemini AI when I asked it to write a short joke about AI and wine. I understand the crack about Tesla drivers, I mean that’s obvious but how

much does it really know that’s helpful to us about tasting wine? A human wineshow judge may write the common line of “overoaked and developed” whereas the AI computer writes

“notes of oak, cherry, and impending obsolescence”. Is the computer just mining our text for insights or can its machine learning techniques be used to create tasting notes outside of its data set? 

In 2025. AI is across all of our lips but how much of it can influence what goes over our lips with wine? Certainly, cyber AI technology can help us bring together “old” technology in the vineyard like weather and soil sensors, GPS, satellites etc to assist in precision viticulture with new technology like sap flow sensors to address the “long term” problem of climate

change and not the “short term” hiccups of heatwaves and droughts. AI can use Blockchain technology to prove the origin and authenticity of premium wines. Blockchain is a method of

recording data and transactions, so it is another layer of transparency to counter fake wine in the ultra-premium end. AI can recommend wines in the bottle shop that suit your palate based on reviews and profile information via apps like Vinvino. But can it match a wine to an estate or a vintage? Maybe yes.

A test has been done that analysed the chemical composition of 80 Bordeaux wines from 12 vintages. The wines came from 7 different chateaux. Machine learning was used to find patterns in the data for 73 of the wines, identifying them as coming from particular Bordeaux estates. Once the model was trained, it was presented with gas chromatograms for the seven held back wines and it was able to assign, correctly, each of them to a particular chateau. The patterns found also a clear separation between left and right bank Bordeaux, and some evidence of vintage differences. So, AI does provide proof that the concept of terroir and winemaking really does have a chemical signal for individual wines or producer, irrespective of vintage. In the future, we may be able to replace blind tasting with machine learning…?

The following two wines each received 5 out of 5 gigabytes and my AI sommelier suggests aging them in repurposed server racks…Happy drinking!