Build me a shortlist of wedding venues to contact. Be honest about what you actually know: venues change names, close, and change pricing, so everything you list is a lead to verify, not a fact.
My wedding: {city}, {month}, {guests} guests, total budget ${budget}.
Preferences:
– Style: {venue_style}
– Must have: {must_haves}
– Do not want: {do_not_want}
Rules:
1. If you can search the web, search, and say that you did. If you cannot, say you are working from memory that may be out of date.
2. Only list venues you are confident actually exist in or near my city. If you are unsure whether one still operates, put it under Verify first. Never invent a venue.
3. State rough capacity and whether it fits {guests}. If you do not know a capacity or a price, write unknown. Never invent numbers.
Format the answer exactly like this, nothing before or after:
1. One table, columns: Venue, Area, Capacity vs my guest count, Style, What to verify.
2. A short list titled Verify first, for anything you are less certain about.
3. Three lines: the exact searches I should run myself. The one question to ask every venue on the first call. The reminder that when quotes arrive, the next step is the Ask the Venue for Its Real Numbers prompt.
Open your answer by stating whether this list came from live search or from memory.
Prompt from weddingish.com/ai, where the current version lives.
What this can and cannot know
A model with live web search, which ChatGPT and Grok can both do, gives you a genuinely current list. One working from memory gives you leads that were true when it was trained, some of which have closed or renamed. The prompt forces it to tell you which kind of list you got. Either way, check every name on a map before you email anyone: this prompt saves you the first hour of searching, not the verifying.
What to change
The three preference fields. Be blunt in the do-not-want field; it filters harder than the wish list. If you leave them blank, sensible defaults fill in and the list will still be useful.
What a good answer looks like
It opens by declaring its source, the table has unknowns in it, and the Verify first list is not empty. A confident list with no unknowns and no caveats is the suspicious one. The per-venue verify column should name specifics: confirm capacity, confirm it hosts ceremonies, confirm it is still operating.
What the model usually gets wrong here
It invents venues, renames real ones, and states capacities with confidence it does not have. It also pads the list with hotel ballrooms no matter what you excluded. Hold it to the rules: unknown is an acceptable answer, an invented number is not.
