Most wine is bought by people standing at a shelf who recognise a label or spot a price and move on. That is not a lack of interest. It is fifty bottles, little time, and nobody around to ask.
Most prompts I write about assume a conversation you have at home. This one is about the two minutes where it actually matters.
What you want
Photo of the shelf, one bottle out of it, with a reason you understand. No list, no jargon, and nothing about bottles that aren’t there.
The hard part is that a photo of a shelf is badly legible. Labels sit at angles, half of each one disappears behind a price tag, and the letters the model does see resemble a hundred other names. That is precisely the situation where a model starts filling in.
The prompt
This is a photo of a shelf in a supermarket.
First, name only the bottles whose label you can genuinely read.
If you are unsure about a name, leave it out. Do not guess.
Then pick one bottle from that list. I'm eating [dish] and want to spend
around [amount]. Say why that bottle and why not the one next to it.
If nothing usable is legible, say so, and tell me what to photograph.
The order is the whole recipe. Read first, choose second. Ask for a choice straight away and reading and choosing become one step, which leaves no moment where the model can admit it cannot see something.
Taken apart
Separate observing from judging. This is the same order as tasting itself: look first, words after. For a model it matters more, because nothing stops it from saying something about an unreadable label anyway.
Give it a way out. “If nothing is legible, say so” sounds redundant and is the most important of the four lines. Without an exit the answer is always a bottle, even when the photo is useless.
Ask about the one next to it. A choice without an alternative is not a choice. Asking why not the neighbour forces a comparison, and that comparison is the only part of the answer you actually learn from.
Where it goes wrong, and how to spot it
Three failures are predictable, and all three are recognisable without knowing the wine.
Names that almost exist. A half-visible label produces a name that sounds like a producer and isn’t one. Always check one name from the list against what is actually in your photo. If that one is wrong, the rest is too.
The wrong price. On a shelf the price tag belongs to the slot, not to the bottle someone put back in it. A model links whatever sits together visually. Ask for the price alongside each name and you see immediately whether the pairing has slipped.
A choice from something that wasn’t legible. The dangerous one. If step one produces a short list and the chosen bottle isn’t on it, the model merged the two steps. Start again with a sharper photo.
Ethics note
A supermarket shelf is not neutral. What sits at eye level is there because someone paid for it, and a photo of that shelf passes that arrangement to the model as if it were the world. So you are not asking “what is the best wine”, you are asking “what is the best of this buying decision”.
Watch the people in your photo too. Someone is always walking through the frame in a shop, and they did not ask to end up in a prompt. Wait a moment, or shoot close.
And the third: this makes buying easier, and easier buying is not the goal. A rule worth adding: let the model ask whether you need wine today at all. That is an annoying question at the shelf and roughly the most useful one there is.
Take it
This is a photo of [what you photographed].
First name only what you can read with certainty. If you are unsure,
leave it out. Do not guess.
Then pick one [thing] from that list for [your situation and budget].
Say why that one and why not the one next to it.
If nothing usable is legible, say so, and tell me what to photograph.
A small exercise: photograph a shelf you know well and see whether the model names something that isn’t there. On a subject where you know the answers, thirty seconds tells you how much to trust the rest.
