These brainstorming prompts get more and better ideas out of ChatGPT, Claude or Gemini: they generate a large list fast, push past the obvious first answers, borrow from other fields, then cluster, score and pressure-test the best ideas. They work for product features, marketing campaigns, names, event plans, research questions and any problem where you need options before you decide.
How to use these prompts
Each prompt is ready to copy into any chat assistant. Replace the blanks in double braces, or keep the default after the colon: {{count:30}} means 30 ideas unless you change it.
Three habits matter more than the wording:
- Give context and constraints. Who it's for, what you've already tried, budget, time, what's off the table. Anthropic's prompting best practices note that explaining the motivation behind an instruction "can give Claude more information about your goals and help it deliver more targeted responses."
- Ask for more than the default. The same guide says: "If you want Claude to do more than the task strictly requires, explicitly request it rather than relying on the model to infer this from vague prompts." Name a number of ideas and ask for range.
- Generate, then judge. Keep the two in separate prompts. Judging while generating tends to trim the unusual ideas first, and those can be the ones worth keeping.
Brainstorming is also one place where running the same prompt twice is useful. OpenAI's ChatGPT prompting guide puts it simply: "Prompting is iterative (and that's normal)." Regenerate, or open a fresh chat, and merge the lists.
Which prompts generate a lot of ideas fast?
1. Wide first list
The default brainstorming prompt. Categories stop the list from being 30 versions of one idea.
I need ideas for {{goal:getting more people to try our budgeting app}}.
Context: {{context:small team, no paid ads budget, users are mostly students}}
Already tried: {{tried:nothing yet}}
Off the table: {{off_limits:nothing}}
Give me {{count:30}} ideas. Spread them across at least 5 different categories and label each category. One line per idea. Don't judge or rank them yet, and include some that seem odd or risky.
2. Ideas under a strange constraint
Constraints force new angles. The constraint doesn't have to be real.
Brainstorm {{count:10}} ideas for {{goal}} under each of these constraints, one at a time:
1. We have zero budget.
2. It must be done in one day.
3. It must work without a website or app.
4. {{custom_constraint:Our audience is over 65}}.
After the lists, pick the 3 ideas that could still work without the constraint, and say why.
3. Ideas from different points of view
Different people solve the same problem differently.
Problem: {{problem}}
Generate 5 ideas each from the point of view of:
- {{perspective_1:a frustrated first-time customer}}
- {{perspective_2:a competitor trying to beat us}}
- {{perspective_3:a teenager}}
- {{perspective_4:someone from a completely different industry}}
- {{perspective_5:our most loyal customer}}
Keep each idea to one line, and mark any idea that showed up from more than one point of view.
4. SCAMPER on something that exists
SCAMPER is a classic checklist for changing an existing product, process or plan: substitute, combine, adapt, modify, put to another use, eliminate, reverse.
Apply the SCAMPER method to {{thing:our weekly team meeting}}.
For each letter (Substitute, Combine, Adapt, Modify, Put to another use, Eliminate, Reverse), give 3 specific ideas. Make them concrete enough to try next week, not general advice. Then mark the 3 ideas that would change the most for the least effort.
Which prompts push past the obvious ideas?
5. Rule out the obvious first
Getting the predictable ideas out of the way lets the model go further.
Topic: {{goal}}
Step 1: List the 10 most obvious ideas anyone would suggest.
Step 2: Now give me {{count:15}} ideas that are NOT on that list and are not small variations of them. At least 5 should feel surprising.
Step 3: For the 3 most surprising ideas, say what would have to be true for them to work.
6. Reverse brainstorm
Asking how to make a problem worse is often easier than asking how to fix it, and the answers flip into fixes.
Goal: {{goal:make new employees productive in their first month}}
1. List {{count:15}} ways we could make this as bad as possible. Be specific and honest.
2. For each one, write the opposite as a practical idea.
3. Mark any "bad" idea that we might be doing already without noticing.
7. Borrow from other industries
Many good ideas are ideas from somewhere else, moved.
Problem: {{problem:customers forget to reorder our product}}
Show how {{count:6}} unrelated industries or fields solve a similar problem, for example {{examples:airlines, video games, hospitals, restaurants}}. For each, name the specific practice and then adapt it into one concrete idea for us. Don't invent facts about real companies; describe common practices only.
8. Collide two ideas
Combining unrelated things is a quick way to get ideas that aren't on everyone else's list.
Here is my list of ideas for {{goal}}:
{{ideas}}
Pick 8 random pairs of ideas from the list and combine each pair into one new idea. Then add 4 combinations of an idea from my list with something unrelated: {{random_items:a subscription box, a podcast, a public library, a board game}}. One or two lines each.
Which prompts sort and choose the best ideas?
9. Cluster a long list into themes
For a long list from an AI, a team session or a survey.
Here is a list of ideas:
{{ideas}}
Group them into themes. For each theme: a short name, a one-sentence description, and the ideas that belong to it. Merge near-duplicates and say which you merged. List any idea that doesn't fit a theme separately; it might be the most original one.
10. Score ideas against your criteria
Your criteria, not the model's. Effort and impact are a start; add what matters to you.
Score these ideas for {{goal}}:
{{ideas}}
Criteria: {{criteria:impact on the goal, effort, cost, time to first result, risk}}
Our situation: {{context}}
Make a table with a 1 to 5 score for each criterion and a one-line reason for each score. Then list the top 3 overall and one "long shot" worth keeping. Say where your scoring depends on information I didn't give you.
11. Pre-mortem the favorite
Imagine it failed and work out why, before you commit time to it.
We're leaning toward this idea: {{idea}}
Imagine it's {{timeframe:six months}} from now and the idea failed. Write the 7 most likely reasons it failed, from most to least likely. For each reason, give an early warning sign we could watch for and one change that would reduce the risk now.
12. Turn an idea into a cheap test
The best way to judge an idea is to try a small version of it.
Idea: {{idea}}
Design the smallest test that would tell us if this idea is worth pursuing. Include: what we'd build or do, how long it takes ({{max_time:under two weeks}}), what it costs, what result would mean "go" and what would mean "stop", and what we'd learn even if it fails.
Which prompts help a team brainstorm together?
13. Plan a brainstorming session
For a meeting where you want everyone's ideas, not only the loudest voices.
Plan a {{length:60-minute}} brainstorming session for {{group_size:8}} people on {{topic}}. It will be {{format:remote, using a shared whiteboard}}.
Include: a short warm-up, a silent idea-writing round before discussion, a way to group ideas, a fair way to vote, and how we leave with owners and next steps. Give timings for each part and the exact questions I should ask the group.
14. Make sense of the sticky notes
Paste the raw notes after the session. Remove names if the notes include them.
Here are the raw notes from our brainstorming session on {{topic}}:
{{notes}}
1. Group them into themes and count how many notes are in each.
2. Merge duplicates.
3. List the ideas that came up most often and the ones that came up only once but seem promising.
4. Write 3 questions the group should answer before choosing.
Keep the original wording of each idea where you can.
How do you get better ideas from AI, not just more?
Quantity is easy; quality comes from what you add. A few things help:
- Feed it real material. Customer complaints, survey answers, competitor pages you've read, your own half-formed notes. Ideas grounded in real problems beat ideas grounded in the model's general knowledge.
- Show an example of a good idea. The Claude prompting guide calls examples "one of the most reliable ways to steer Claude's output format, tone, and structure." Paste one idea you like and say why.
- Iterate on the best few. Take three ideas from the first list and ask for ten variations of each.
- Check facts before you act. If an idea depends on a claim about a market, a law or a real company, verify it from a primary source. Anthropic's guide to reducing hallucinations is direct about the limits: its techniques significantly reduce hallucinations, but they "don't eliminate them entirely."
For prompts that make the model reason step by step when judging ideas, see the guide to chain-of-thought prompting.
Where should you keep your best brainstorming prompts?
The prompts above become more useful once they carry your defaults: your product, your audience, the constraints you always work under. Save a version for each kind of brainstorm you run often, such as campaign ideas, feature ideas or names. For idea work on a new business, the startup prompts go further.
Kipt keeps prompts like these as Markdown files in a private GitHub repo you own. Blanks like {{goal}} stay fillable, tags group your brainstorming prompts, and ⌘K finds the reverse brainstorm when a meeting stalls.