What's the fastest way to get unstuck when writing a novel?

It depends which kind of stuck you have. If you know what happens next and the sentences won't come, a generation tool that drafts the passage gives you something to react to, which is usually faster than staring. If you don't know what happens next, generated prose tends to make it worse — you get competent text going in a direction you never chose. That's the point to stop producing and start looking at the premise.

Should I let AI write part of my novel?

That's yours to decide, and writers land in different places on it. The practical note is narrower — text you didn't choose the direction of takes longer to fix than to write, so generation helps most when you already know where the scene is going and least when you don't.

How is "watching a story play out" different from asking ChatGPT what happens next?

Asking a chatbot gives you a summary of possibilities — a list of options described from outside. Reading it play out means the scene actually happens, in order, with the consequences following, and you keep steering as it goes. One tells you about the story; the other runs it. Which is more useful depends on whether you need options or need to see it.

Doesn't using AI at all make the writing generic?

The common failure is handing over the deciding, not the typing. Prose you generated and then rewrote in your own voice reads like you. Prose you generated because you didn't know what should happen tends to read like nothing in particular — because no one chose it.

I'm not writing a novel. I just like stories.

Then the second half of this is the whole thing, without the writing part attached. Putting in a situation and reading where it goes is a way to spend an evening, not only a technique for a stuck manuscript.

Stuck on a story — when to have AI write, and when to watch it play out

2026-09-11 · PLOT Team

TL;DR

Two different problems get called writer's block — missing sentences, and a premise you cannot see yet.
Generation tools solve the first one. They hand you prose to push against.
For the second, asking "what if X happened?" and reading the answer is a different move from asking for text.
Neither replaces deciding what the story is. That part stays yours.

Type "AI for writer's block" into a search bar and the results agree with each other. You get tools that generate — describe this scene, expand this paragraph, draft this chapter. The premise is consistent: you're stuck because there aren't enough words yet, so here are some words.

That's a real answer to a real problem. It just isn't the only kind of stuck.

Two problems with the same name

The first kind is missing sentences. You know what happens. The scene is clear — she finds the letter, she doesn't open it, she puts it back. You just can't get it onto the page in a way that isn't dead on arrival.

This is what generation tools are for, and they're good at it. Something on the page you disagree with is easier to fix than an empty page, because disagreement is specific. You read the generated version, you know immediately that she wouldn't hesitate that long, and now you're writing.

The second kind is not knowing what happens. The chapter ends and you have no idea what the next one is. You have a premise, a couple of characters, a feeling about the tone — and no events.

Generating prose here does something quietly bad. It fills the gap with plausible text going in a direction you never chose, and plausible-but-unchosen is harder to fix than blank, because it looks finished. You end up editing a scene you don't believe in instead of finding the one you do.

The move for the second kind

When the problem is that you can't see it yet, the useful question isn't "write this for me." It's "what if this happened?"

Most advice about AI and stuck stories converges on the same technique — throw a premise at it and see what comes back. Interview a character about something they haven't told you. Ask what breaks if the antagonist simply gives up in chapter four. Push a scene to an extreme and look at the wreckage.

There are two ways to run that question, and they give you different things.

Ask a general chatbot and you get options — three or four described from outside, in summary. He could confess. She could find out from someone else. The letter could be a forgery. Useful when you want a menu.

Read it play out and the scene actually happens instead of being described. She finds out from someone else, and you see what that does to the room, and then the next scene follows from it, and you keep steering as it goes. You're not evaluating a proposal. You're watching a version of the story exist for a few minutes.

That's a different sense than reading a list. A premise can look fine in summary and feel wrong the moment it's in motion — and you generally can't tell which until you see it move.

What comes out, and what doesn't

Nothing you get this way is manuscript material, and it's worth saying that plainly.

The prose isn't yours, the pacing is the engine's, and none of it is structured as a draft. What you take away is smaller and harder to point at: you now know that the confession lands flat, that the sister is more interesting than the rival, that the story wants to be about the house and not the inheritance. Then you close it and write the chapter yourself.

Which is the actual boundary. The looking is not the writing. Whatever you learn from watching a premise move, the book still gets written by the person whose name is on it.

In practice

You know what happens, the words won't come. Use a generation tool. Have it draft the passage, then rewrite it until it sounds like you.
You don't know what happens. Stop producing text. Ask what-if, and watch one version run.
You know what happens and you don't like it. Same as above — run the version you rejected and find out whether you were right.
You're not stuck at all, you just want a story tonight. Then skip the writing frame entirely. That's the whole activity.

The trap is using the first move on the second problem. It's the more available option — the tools are right there, and producing text feels like progress. But text you didn't choose the direction of costs more to remove than it did to make.

Where PLOT fits

PLOT is the second move. You put in a situation in your own words — or just pick a genre — and read what happens, steering at the end of each page or by typing a line. It runs the premise instead of describing it.

It won't draft your chapter. There's no export, no manuscript view, nothing to paste into your document. For the first kind of stuck, use the tools built for that. For the second, it's a way to see the thing before you commit to writing it.

And if you're not writing anything at all, it works exactly the same way. Most people using it aren't.

Frequently Asked Questions

What's the fastest way to get unstuck when writing a novel?

It depends which kind of stuck you have. If you know what happens next and the sentences won't come, a generation tool that drafts the passage gives you something to react to, which is usually faster than staring. If you don't know what happens next, generated prose tends to make it worse — you get competent text going in a direction you never chose. That's the point to stop producing and start looking at the premise.
Should I let AI write part of my novel?

That's yours to decide, and writers land in different places on it. The practical note is narrower — text you didn't choose the direction of takes longer to fix than to write, so generation helps most when you already know where the scene is going and least when you don't.
How is "watching a story play out" different from asking ChatGPT what happens next?

Asking a chatbot gives you a summary of possibilities — a list of options described from outside. Reading it play out means the scene actually happens, in order, with the consequences following, and you keep steering as it goes. One tells you about the story; the other runs it. Which is more useful depends on whether you need options or need to see it.
Doesn't using AI at all make the writing generic?

The common failure is handing over the deciding, not the typing. Prose you generated and then rewrote in your own voice reads like you. Prose you generated because you didn't know what should happen tends to read like nothing in particular — because no one chose it.
I'm not writing a novel. I just like stories.

Then the second half of this is the whole thing, without the writing part attached. Putting in a situation and reading where it goes is a way to spend an evening, not only a technique for a stuck manuscript.

Try PLOT on the web →

Google Play · App Store

Sudowrite vs PLOT — write a story, or read one? · Sudowrite vs NovelCrafter vs Scrivener · An AI story app to read, not a writing tool