Testing a Talk Idea Before You Write It

What this is for

Writing a full talk around an idea that doesn't hold up is expensive. This guide uses AI to pressure-test an idea while it still costs you nothing to change it - before you've built slides, before you've written an opening, before you've told anyone the title.

When to use it

Use this once you have a candidate angle from finding the talk only you can give (guide 10). Run it here before you outline anything. If the idea survives this pass, move on to structure. If it doesn't, go back and find a sharper one. That's a normal part of the process, not a failure.

The prompt kit

I'm considering building a talk around this idea: [describe the idea in 2-3 sentences]

My audience is: [who you speak to]

Before you react to it, ask me three questions that would help you understand whether this idea actually holds up. Don't tell me what you think yet.

Here's what I've told you about the idea: [paste the idea plus your answers from the last prompt]

Now play the skeptic in my audience. What's the strongest objection someone in [audience type] would raise against this idea? What would make them tune out or push back?

I want to know if this idea is actually new, or a version of something [audience type] has already heard a dozen times. Based on what I've told you, name the most common version of this talk that already exists on the circuit. Then tell me straight: is my version different enough, or am I retelling the common one?

Assume this idea survives and I build the talk. Give me the one-line summary of what someone in the audience should be able to say to a colleague the next day. If you can't write a tight one, tell me the idea is still too broad and ask me the question that would narrow it.

What good output looks like

The AI should push back somewhere. If every prompt in this kit comes back with encouragement and no real objection, it isn't testing the idea. It's agreeing with you. A useful pass surfaces at least one real weakness: an objection you hadn't considered, a competing talk you didn't know existed, or a summary line that comes out mushy because the idea itself is still too broad.

Watch out for

Treating disagreement as a personal note. The AI raising a strong objection about your talk idea is not a comment on you as a speaker. Read it as information about the idea, decide if it's fair, and either sharpen the idea or explain why the objection doesn't apply.

Only asking once. One round of questions rarely surfaces the real problem. Run the objection prompt, revise the idea based on what came up, and run it again. Two or three passes is normal.

Confusing agreement with quality. These tools default to agreeable responses unless you explicitly ask for the skeptic's view, as the prompts above do. Skip straight to "what do you think of this idea" and you'll get a warmer answer than the idea deserves.

Letting the AI invent the competing talk. If it names a "common version already on the circuit," ask where that comes from. If it can't point to something specific, treat it as a guess, not a fact, and check it against your own knowledge of the speaking world.