For most of its history, wine discovery has run on two things: famous names and other people's opinions. Artificial intelligence is changing both — not by replacing expertise, but by structuring it.
The problem with crowd-sourced ratings
The dominant model for online wine discovery has been the aggregate rating: millions of users scoring bottles out of five. It's useful for spotting crowd favourites, but it has real limits. Ratings tell you whether other people liked a wine on its own — not whether it suits your meal, your palate, or the dish in front of you.
Worse, popularity compounds. Well-known wines accumulate reviews and rise to the top, while excellent lesser-known bottles stay invisible. Discovery becomes a feedback loop that rewards fame, not fit.
From opinions to structured data
The shift now underway is from opinion to data. Instead of scraping reviews, the most interesting wine platforms are building structured databases from producer technical sheets — the factual documents that describe a wine's grape composition, region, vinification, acidity, tannin structure, residual sugar and serving guidance.
This matters because a machine-learning model is only as good as its inputs. A model trained on real chemical and structural data can reason about why a wine works with a dish. A model trained on star ratings can only tell you that strangers enjoyed it.
Good wine AI doesn't replace the sommelier — it gives every table access to the structured knowledge a great sommelier carries in their head.
Pairing as a computational problem
Once wine is described as structured data, pairing becomes something a model can actually compute. The same five levers a sommelier weighs — acidity, tannin, sweetness, body and intensity — can be represented numerically and matched against the characteristics of a dish, including its cooking method and sauce.
This is the foundation Entwine is built on. Its database of 8,000+ wines comes from producer sources, and its pairing engine matches those wines to the exact dishes on a restaurant's menu. The recommendation isn't a guess drawn from a popularity chart — it's a calculated fit, with a reason behind it.
What this means for diners and restaurants
For diners, it means discovery that finally accounts for context: the right wine for this plate, from the list actually in front of you. For restaurants, it means the intelligence of a seasoned sommelier available at every table, plus data on what's selling, what's being paired and what's missing from the list. The age of the static five-star rating is ending — and something far more useful is taking its place.
See wine intelligence in action.
Entwine pairs 8,000+ wines to real menus, at every table where wine is taken seriously.
Download the app →