Use when you want a poll to inform a decision rather than just collect votes.
poll-that-returns-answers.md
You are designing a social poll as a piece of research. You are not designing it to farm engagement.
The decision this poll should help with: {{DECISION}}
What we currently believe the answer is: {{OUR_HYPOTHESIS}}
Who follows the account and how representative they are: {{FOLLOWER_PROFILE}}
Platform and its option limit: {{PLATFORM_LIMITS}}
Output four numbered sections.
1. Three poll versions. For each: the question under 100 characters, the options within {{PLATFORM_LIMITS}}, and one line per option on what its winning would tell us about {{DECISION}}. If an option would not change what we do either way, drop it and say why.
2. Bias check: for each version, the leading wording, the option you left out, and the answer people give to look good rather than truthfully. Rewrite the worst offender in full.
3. The pinned follow-up comment asking voters for the reasoning behind their vote, since the reasons are the part worth reading.
4. Interpretation limits: what {{FOLLOWER_PROFILE}} makes this poll unable to tell us, and the response count below which we should not act on the result.
Test every version against {{OUR_HYPOTHESIS}}. If a version cannot return a result that contradicts it, the poll is decoration and you should say so. Do not offer "other" as an option unless the follow-up comment collects what other meant. No em dashes.
Replace each placeholder with your own detail. The more specific you are, the less the model invents.
When should I use this rather than just posting a poll?
When the poll is supposed to inform a decision. If you want reach, a poll works as a post and needs none of this structure. Reach for it when a result would genuinely change what you build or write, since section one drops any option whose winning would leave you doing the same thing.
What do I need in front of me?
The decision it should help with, your current hypothesis, an honest description of who follows the account, and the platform option limit. The hypothesis matters most: without it the prompt cannot check whether a version is capable of returning a result that contradicts what you already believe.
What comes back, and where is the real data?
Three poll versions with what each result would decide, a bias check naming the leading wording and the option left out, a pinned follow-up comment, and interpretation limits with a minimum response count. The replies under that pinned comment are the data worth reading; the votes mostly recruit people to write them.
What is the mistake?
Presenting a follower poll as market research. The sample is people who already chose to follow you, so it tells you about your followers and not about your market. How much weight social polls deserve is argued in marketing circles, but the sampling problem is not really in dispute.