How Sommo Learns Your Palate
Vivino tells you what the crowd thinks. Sommo AI learns what you think, and notices when that changes. Here is how your own journal becomes a living profile of the wine you actually like.
Step by step.
You log what you taste
Scan or add a wine, give it a three-tier rating, and add notes if you feel like it.
Your profile refreshes
After a few entries, the app synthesises the patterns hiding in your history.
Recommendations get personal
From there the app leans on your taste, not a crowd average.
Most wine apps answer “what should I drink?” with a crowd score. Sommo AI answers it with you. The longer you use it, the better it knows your palate, not because it watched what strangers rated, but because it read your own journal. This is the quiet engine under the whole app, and it is built entirely from wines you have actually tasted. It is also the feature I am proudest of, so let me explain why I built it the way I did.
The loop
Everything turns in a small loop:
- You scan or add a wine, and rate it.
- The entry lands in your journal.
- Once you have enough entries, your profile refreshes.
- Sommo AI uses that profile to recommend, pair, and prioritise.
Then you taste the next wine and the loop tightens. Nothing here depends on a personality quiz or a sign-up survey. The only input is the wine in front of you.
Why this beats a crowd score
I believe, genuinely, that every palate is unique. A glance at the world makes the case for me: someone who grew up eating in Sichuan, someone raised on French home cooking, and someone weaned on a Turkish table do not arrive at wine with the same mouth. Different food, different spice, different memories of sweetness and bitterness and salt. Of course the wines that move them differ too.
A crowd score flattens all of that into a single average. It is the taste of nobody in particular: useful for spotting a flawed bottle, useless for telling you whether you will enjoy it. Sommo AI starts from the opposite end. It does not ask “what does everyone think?”. It asks “what do you think?”, and it only has one source for that answer: your own journal. That is the whole difference between a popularity vote and a profile of your actual taste.
Your palate is not fixed
Here is the part I am surprised almost no one builds for. Taste is not static, and it is not linear. We change. Our mouths change.
I used to hate tea. Ten years ago I would not touch it, and now I reach for it over coffee most mornings. I like big, warming reds in the depth of winter and barely glance at them in July. Films I did not understand at twenty land completely differently at thirties, not because the film changed, but because I did. Change is not a bug in a person; it is the most human thing about us.
So why would a wine profile freeze you on the day you signed up? Sommo AI is built to notice palates drift and to reason about the pattern, not just average everything you have ever logged. The mechanism is deliberate: recent entries count for more than old ones, on a roughly six-month half-life, so a shift towards lighter wines, or away from heavy oak, or with the turning seasons, actually surfaces. Older entries never vanish entirely, they just quietly matter less, the same way your tastes of a decade ago still shaped you but no longer rule you.
I have not seen another wine app, or honestly many personalisation engines anywhere, treat you as a moving thing rather than a fixed label. As far as I can tell, Sommo may be among the first to think about it from this side. To me it is obvious: a profile that cannot change is a profile that is wrong a little more every month.
Wine character analysis
Once you have at least three entries, Sommo AI can generate a wine character analysis: a short, written personality profile of your palate. It reads your ratings and notes and names the patterns you might not have noticed, the styles you gravitate to, the ones you cool on, the threads that run through your favourites.
It is cached cleverly. The analysis is keyed to the actual content of your journal, so it only regenerates when your entries change, never on a timer. Log a new wine and the next refresh reflects it. Change nothing and it stays put.
Taste DNA
Where the character analysis is prose, Taste DNA is a fingerprint. It turns your journal into a set of axes, each scored on the same one-to-five scale as professional tasting structure, so you can see the shape of your palate at a glance: how much body, tannin, acidity or sweetness you reach for. It carries the same recency weighting described above, plus an adventure score that measures how widely you range across grapes and regions, so the profile can tell a happy specialist from a restless explorer.
Three-tier ratings
At home, a 100-point scale is false precision. Could you honestly tell an 89 from a 91 across a year of bottles? Sommo asks the only question that matters in the moment:
- Not for me.
- Solid choice.
- Would buy again.
Three honest taps beat a fake number. The simplicity is the point: it is fast enough that you actually do it, every bottle, which is what makes the profile real. Each entry earns a little XP, and a full structured tasting note earns a bonus, because the richer the note, the more it teaches Sommo AI about you.
Structured tasting notes
If you want to go deeper, Sommo guides you through the Structured Approach to Tasting: appearance, nose, palate, conclusions. The tasting notes feature walks the same path a WSET student learns, and the glossary entry on tasting notes explains the vocabulary. Advanced notes both train your own palate and feed the profile more signal, which is why they earn that XP bonus.
What I never do
Your palate is personal, and it stays that way:
- I never sell your journal or your profile.
- I never use your entries to train the model other people use.
- There is no public feed, no followers, no leaderboard. Sommo is a notebook, not a social network.
The full commitments are on the privacy and data page.
Getting the most out of it
- Log honestly, not aspirationally. Rate what you actually felt, including the wines you did not like. Dislikes shape the profile as much as favourites.
- Use structured notes on the bottles that puzzle you. That is where the learning is.
- Open and rate bottles from your cellar. It teaches Sommo AI and, in time, calibrates your drinking windows to your taste.
New to all this? Start with Sommo for wine beginners.
What’s coming
The natural next step is letting your palate quietly adjust your drinking windows, so a wine’s “ready” date bends towards when you tend to enjoy that style. That work is in progress, and I would rather ship it properly than promise it early.
In the app
You will find this across three screens: the journal list of everything you have tasted, the character analysis card, and the Taste DNA fingerprint. All three are built from the same place, your own glass.
Frequently asked.
01.How many wines do I need to log before this works?+
02.How is this different from Vivino's ratings?+
03.What if my taste changes over time?+
04.What happens to my profile if I downgrade from Premium?+
05.Is my journal used to train the AI?+
Try it yourself.
Free to download, with five label scans to start and one Premium subscription that costs less than a bottle of decent wine per month.