Use the details that change the decision, keep real constraints visible and check facts the model cannot verify.
You do not need a magic sentence.
You need a clear task, the context that changes the answer and a useful way to judge what comes back.
Build the prompt
Add what you know. Leave blank what you do not.
A useful gift brief usually has four parts.
1. The situation
- relationship
- occasion
- budget
- deadline or location if they change what is possible
2. The person
Use current, gift-relevant details:
- interests they actually spend time on
- routines
- current-life changes
- things they keep mentioning
- small recurring annoyances
3. The exclusions
- obvious duplicates
- past gifts
- dislikes
- clutter
- sizing
- allergies
- specialist categories you do not understand
- anything inappropriate for the relationship
4. The output you want
Instead of “give me ideas,” ask for something you can inspect:
- a small number of genuinely different directions
- the clue behind each one
- one risk or assumption to verify
- questions first if the information is too thin
- no invented live prices, stock or delivery
The master prompt
Help me choose a gift for one specific person. Relationship: [relationship] Occasion: [occasion] Budget: [budget] Deadline/location if relevant: [deadline/location] Useful things I know about them: [current interests, routines, life changes, repeated comments] They already own / I already gave: [history] Avoid: [dislikes, duplicates, sizing, allergies, clutter, social boundaries] Give me 6 gift directions that are genuinely different from one another. For each, tell me which detail from my description led you there, why it might fit, and one reason it could miss. Do not invent live prices, stock or delivery. If the information is too vague to do this well, ask me up to 3 questions first.
The structure matters more than the exact wording.
Use a specialist prompt when the problem is specific
- Birthday Gift Prompt: the birthday itself changes the brief.
- Christmas Gift Prompt: several people need to stay separate.
- Someone Who Has Everything: normal product ideation is the wrong objective.
- Someone You Barely Know: uncertainty and social fit matter.
- Birthday Tomorrow: the deadline removes most normal options.
Do not force every edge case into one giant prompt.
Repair the first answer instead of starting over
Too repetitive
Keep the facts I gave you. Give me three new directions that do not repeat any category from your first answer.
One hobby took over
Keep [hobby] as one possible direction, but use the rest of the person for the others.
Too generic
Remove any idea that would fit a random person with the same age or gender. Tie every remaining idea to a clue I actually gave you.
Invented traits
Use only the facts I provided. Mark assumptions and ask me questions where the missing information matters.
Shopping facts look made up
Separate gift directions from live shopping facts. If you cannot verify current price, stock or delivery, tell me what to search for instead.
ChatGPT, Gemini or Claude?
You usually do not need three completely different recipient briefs.
The same core structure works: direct instruction, relevant context, explicit constraints and a useful output format.
The exact response will still vary by product, model and version.
Treat the prompt as a brief, not a spell.
What the prompt cannot verify by itself
Before buying, check whatever is current or personal:
- price
- stock
- delivery
- seller quality
- sizing
- allergies
- booking conditions
- whether they already bought it
- whether the gift feels right for your relationship
The prompt narrows the thinking. It does not turn a plausible suggestion into a verified purchase.
