Prompts For Brainstorming And Idea Generation

AI is a genuinely useful thinking partner for generative work — but only if you prompt it correctly. The default response to a vague brainstorming request is a list of obvious, generic ideas that you could have thought of yourself in two minutes.

Getting ideas that are actually useful requires prompts that push the model beyond the first layer of obvious associations — and techniques that use AI as a thinking partner rather than just an idea dispenser.

Why Default Brainstorming Prompts Fail

When you type give me ideas for [topic], the model produces what is statistically most common in its training data — which is, almost by definition, what everyone else has already thought of.

The ideas that feel obvious to you will feel obvious to AI for the same reason: they're the ones most frequently associated with the topic in text.

To get past this, your prompt needs to do one or more of the following:

• Constrain the idea space in an unusual way — forcing the model to look at the problem from a different angle;
• Specify the type of idea you're looking for — conventional, unconventional, contrarian, or high-risk;
• Add friction deliberately — ask for ideas that are hard to execute, ideas that would be rejected by most people, or ideas that solve a related but different problem;
• Use iteration — ask for a first batch, then push further by asking for alternatives to the ones that feel too safe.

Core Brainstorming Prompt Templates

For generating a range of ideas:

For pushing beyond the obvious:

For structured divergent thinking:

Using AI As A Thinking Partner, Not Just An Idea Generator

The most productive brainstorming sessions with AI aren't one-shot requests — they're dialogues. You can use the model to push your own thinking in ways that go beyond generating a list:

Challenge your assumptions: Here is our current approach to [problem]. What are the three biggest assumptions baked into this approach? What would happen if each one turned out to be wrong?

Generate objections: Here is an idea I'm considering: [describe idea]. Act as a skeptic who wants to find reasons this won't work. Give me the five strongest objections to this idea.

Force combinations: Here are two unrelated ideas: [idea A] and [idea B]. Generate three ways to combine elements of both into a single solution for [problem].

Reframe the problem: I've been trying to solve [problem] by [current approach]. What if I'm solving the wrong problem? What related or underlying problem might actually be worth solving instead?

Practice: From Generic To Genuinely Useful

Pick any problem or creative challenge you're currently working on. Run through this sequence:

1. Start with a standard brainstorm prompt — give me [number] ideas for [topic] — and collect the output;
2. Identify the two or three ideas that feel most generic or predictable;
3. Send a follow-up that explicitly excludes that category of thinking and asks for a different angle;
4. From the second round, pick the most interesting idea and ask AI to generate three objections to it.

After this sequence, you'll have a set of ideas that have been pushed further, stress-tested, and refined — in under ten minutes.