How to use LLMs to pick wine

For me, buying wine is partially narrowing down the choices to wines I tend to like, and then tie-break based on price and vibes.

I find that ChatGPT is actually pretty good at automating this for me. I take a picture and paste in the following prompt, and have been impressed by the results:

You are a wine sommelier's assistant. In the attached image you’ll either see a shelf of wine or a wine menu.

1. **Identify** each visible bottle’s:
   - Name (brand and label)
   - Grape variety
   - Region (and vintage, if readable)
   - Approximate price (if visible on the label or shelf tag)

2. **Based on my taste profile and value criteria**, rank your **top 3** picks that cost **$20 or less** and explain each choice in 2–3 sentences, focusing on:
   - Dryness level (I like more on the dry side)
   - Spice/tannin intensity (I enjoy moderate to pronounced tannins and spicy notes, like a Malbec)
   - Flavor profile (I favor berry and plum notes)
   - Value (why it’s a standout at or below $20)

3. **Bonus**: If none perfectly match, suggest **one** outstanding out-of-my-comfort-zone option under $20 and explain why it’s exceptional.

Keep your answers concise and practical. If a label or price is unreadable, note that and move on.

I also did a similar thing for picking a bottle of whiskey in a hurry and it did an excellent job-- it picked a well-priced bottle for a California-based distillery that I had never heard of.

Copyright Ricardo Decal. richarddecal.com