Anyone using AI tools for outlining/brainstorming only?

@plotholes_anon OP

2mo ago

I'm firmly in the camp that the actual prose has to be mine, word for word, no exceptions. But I get stuck on plot logic and I'm wondering if AI is useful purely as a brainstorming partner: poking holes in my outline, listing what could go wrong, that kind of thing. Anyone using it that way without letting it touch the writing?

11 comments

@plotholes_anonOP2mo ago

Yes, and the framing that keeps it healthy is rubber-duck, not co-author. I paste my premise and ask it to interrogate the logic: why wouldn't the villain just do X, what's stopping the protagonist from leaving in chapter one. It's genuinely good at surfacing the question I've been subconsciously avoiding. Then I close the tab and write the answer in my own words.

@the_synopsis_struggle1mo ago

The single prompt that earns its keep for me: 'list twenty things that could go wrong for this character given this goal and this setting.' I throw out fifteen, two are obvious, three crack the act-two sag wide open. It's a divergent-thinking machine. The choosing and the writing stay entirely mine.

@chapter_two_curse1mo ago

The 'what could go wrong' list is brilliant for the muddy middle. Do you give it your characters and setting first, or keep it abstract so it doesn't start writing scenes at you?

@the_synopsis_struggle1mo ago

Give it the setup but explicitly tell it 'bullet points only, no prose, no dialogue.' The instant it starts drafting actual scenes you're on the slope you're trying to avoid. I literally end prompts with 'do not write any narrative.' Keeps it in the brainstorming lane.

@plotholes_anonOP1mo ago

This is the nuanced take I came for. The consensus seems to be: outline and logic interrogation yes, lists of what-could-go-wrong yes, premise-poking yes, and the prose stays human, full stop. The 'bullet points only, no narrative' instruction is the guardrail I was missing. Thanks, everyone.

@penname_pending1mo ago

It's a fantastic interrogator of your premise. I ask it to argue against my story's central assumption, like 'convince me this character's choice in chapter ten is unbelievable.' Half the time it's wrong and I feel reassured, half the time it's right and I just saved myself a beta reader's worth of that note.

@wattpad_to_print1mo ago

A boundary I had to learn the hard way: don't paste your actual prose in asking for 'feedback on the writing,' because the suggestions creep into your voice even when you retype them. Keep the AI strictly at the outline and logic layer. Sentences and scenes are a different room it doesn't get into.

@brewed_prose1mo ago

Worth being clear-eyed about two limits. One, it confidently invents 'facts' about story structure, so treat its claims as prompts to think, not rules. Two, on the legal and ethical side, brainstorming is a grey area many of us are fine with, but anything resembling generated prose raises real authorship and rights questions, especially if you publish. The further it stays from the page, the cleaner you are.

@once_more_writer1mo ago

Where it's saved me is as a structure checker, not an idea generator. I paste my own beat sheet and ask which beats are missing or which character's motivation goes dark for too long. It's spotted a flat midpoint for me more than once. Critiquing my outline feels categorically different from writing it, and I'm at peace with that line.

@half_finished_hugo1mo ago

My low-tech version doesn't even need a fancy model: I keep a running notes doc and use the AI like an improv partner that always says 'yes, and what if.' The energy of being asked twenty questions about my own world unsticks me. Then I switch apps entirely to write, so there's a physical wall between brainstorming and drafting.

@query_trenches1mo ago

One more use that stays safely off the page: untangling my own logline and synopsis. I describe the plot in a messy paragraph and ask it to mirror back what it thinks the core conflict is. When its summary is wrong, that tells me my story isn't communicating what I think it is. Diagnostic, not generative.