The second-cheapest wine on any list has the worst markup. You've probably been ordering it. Somm AI analyzes any restaurant wine list by URL and returns your top 10 value picks β every bottle scored on quality, retail markup, regional value, vintage, and market dynamics. Free. No account. Takes 10 seconds.
No reviews yetBe the first to leave a review for Somm AI
Maker
π
Hey PH β I'm Jeremy, solo founder of Somm AI.
I built this because I kept watching smart people freeze when the wine list arrived. They'd either order the second-cheapest bottle (usually the worst markup on the list) or drop $90 on something that sells for $30 at retail.
Sommeliers aren't wizards. They just have a framework: read the list's structure, find the overlooked sections, spot where pricing psychology breaks down. Somm AI applies that framework to any list in seconds.
Paste in a wine list URL, set your budget and style, and you get a ranked top 10 with reasoning β not "this is a good wine," but why it's the best value on this list tonight.
Three things I want feedback on:
Use cases beyond the dinner table β people are already using it for corporate events and wine bar crawls
The output β is the reasoning useful, or do you just want the picks?
Where it breaks β wine lists come in every format imaginable, so tell me where it fails
Free at aisomm.io. Tell me what you find.
Report
Does it pull pricing data from somewhere specific like Wine-Searcher, or are you building your own benchmark for what counts as a fair retail price?
Report
Maker
@hkalmuk14663Β Great question β neither, actually, and that's deliberate (for now).
Right now Somm-AI doesn't lean on an external retail price feed. Value is reconstructed structurally: producer track record, vintage quality for the region, and appellation classification give each bottle a quality signal, and then every wine is compared against its peers on that specific list. So a bottle isn't "good value" against an abstract retail number β it's good value relative to what else you could order tonight at that restaurant, which is the decision a diner is actually making.
The list-relative approach also means it works on any wine list, including obscure producers and older vintages where retail feeds get thin or stale.
That said β a retail markup layer is on the roadmap. When it ships, it'll be a benchmark blended from multiple market sources rather than a single feed, used as an additional signal on top of the structural model rather than the foundation. Happy to go deeper on the scoring model if you're curious.
Report
Tested it on a place near me and it flagged a bottle I've walked past for years that was way better than what I usually grab. The score breakdown actually makes sense too.
Does it pull pricing data from somewhere specific like Wine-Searcher, or are you building your own benchmark for what counts as a fair retail price?
@hkalmuk14663Β
Great question β neither, actually, and that's deliberate (for now).
Right now Somm-AI doesn't lean on an external retail price feed. Value is reconstructed structurally: producer track record, vintage quality for the region, and appellation classification give each bottle a quality signal, and then every wine is compared against its peers on that specific list. So a bottle isn't "good value" against an abstract retail number β it's good value relative to what else you could order tonight at that restaurant, which is the decision a diner is actually making.
The list-relative approach also means it works on any wine list, including obscure producers and older vintages where retail feeds get thin or stale.
That said β a retail markup layer is on the roadmap. When it ships, it'll be a benchmark blended from multiple market sources rather than a single feed, used as an additional signal on top of the structural model rather than the foundation. Happy to go deeper on the scoring model if you're curious.
Tested it on a place near me and it flagged a bottle I've walked past for years that was way better than what I usually grab. The score breakdown actually makes sense too.