AI Brainstorming: Generate Alternatives You Can Actually Compare
A long list of ideas can create more sorting work than insight. Give brainstorming a decision structure: a specific problem, real constraints and contrasting approaches. The useful output is a small set of genuinely different options with enough detail to test their assumptions.
Why it matters
Without constraints, many suggestions converge on familiar generic answers. Naming the budget, audience and resources encourages options that fit the situation rather than imaginary teams with unlimited time.
A practical workflow
- Describe the problem in observable terms and explain why existing attempts have not solved it.
- Set a few real constraints, including available time, skills and resources that cannot be assumed.
- Request options built around different mechanisms, with a main assumption and a cheap test for each.
- Choose one experiment based on what you can learn quickly, rather than which idea has the strongest name.
An example
A student club wants more workshop attendance. Ask for distinct approaches based on scheduling, topic selection and partnerships. Each option should identify a test, such as polling preferred times or running one joint session, rather than repeating 'post more on social media'.
A mistake to avoid
Do not confuse novelty with feasibility. A surprising idea may depend on permissions, equipment or relationships you do not have. Ask which assumption would invalidate it before investing in production.
Try it and check the result
Run one small experiment and record what changed. Bring that result into the next brainstorming round so the model works with evidence from your audience instead of restarting from generic possibilities.
Reference and further reading
Google's prompt design strategies.
The workflows and examples in this guide are illustrative. Product controls and availability can change; check the linked official guidance for your setup.
Continue with Write Better AI Prompts with a Simple Task Brief or explore all blog topics.