How to Use AI as a Screenwriter: Three Simple But Effective Methods

Introduction

AI has proven that it is here to stay.*

While it’s busy taking all the entry-level white collar jobs, at least things like screenwriting are safe for the moment. No one is expecting AI to be able to produce a competent screenplay all by itself, even if you go to all the trouble of scaffolding scenes, structure, etc.**

That is, at least until people are dumbed down enough not to be able to tell the difference between AI slop and an actual human-written screenplay. And, believe me, that’s the plan of the Money People–so unless something dramatic changes soon, don’t expect screenwriting as such to be a viable career path.

But that’s neither here nor there, is it? You, like I, write because we don’t really have a choice in the matter. We actually feel compelled by the process. It feels psychologically healthy to get the creative juices flowing and actually scrawl on a blank page. We are also megalomaniacal enough to think someone else would want to read it.

Not to mention, as musician Ray Harmony suggests, even if AI makes things easier, it’s not that we create because to do so is easy. In some sense, we specifically create art because it is not easy.

With that in mind, it doesn’t seem like incorporating AI would be a particularly good idea as a writer. After all, every query you’re using and every data point you give the machine is just giving the golem in the cloud more fodder for its hopper. Essentially, by using AI, you’re just training the thing that’s trying to take your job.

Be that as it may, that’s a cynical viewpoint. A better way to look at it would be whether you can use AI in effective ways that would spur your own creativity. And that’s what I want to discuss here.

Look, no one is expecting the robot to be able to write a competent screenplay.

This might end up being off-label uses–I mean we can’t all be high schoolers cheating on our English papers–but I guarantee these can be used to improve your screenwriting workflow, from the time-consuming and arduous task of writing your supplementary materials to raising the stakes in your logline to core idea generation.

Let’s get started.

*Provided that we can extract enough dinosaur juice from below the surface of the Earth to keep this shit alive, all while surviving all the downstream effects of our one-time fossil fuel energy bonanza.

**And why go to all the trouble when it’s simpler and more fulfilling just to write the fucking screenplay yourself?

Writing Supplementary Materials

There are a couple of ways to go about how to use AI as a screenwriter.

Personally, I like to start with simply writing my own 250-word synopsis and 25-30-word logline.

First, I will ask the AI to pretend that it is “a film development executive critiquing a screenwriter’s pitch” critique these and ask any questions that it might have.

Note these questions down, as they are likely things you want to address in your rewrite.

Second, I give the AI as much information as I have, whether in a longer outline or even a full script. You can ask it to pretend to be “the marketing department of a production company prepping a pitch deck for film festivals,” for example.

At this point, see how closely what the AI says hews to your actual work. Consider what the AI is misunderstanding. Has it filled in any gaps?

Remember, hallucination is a feature, not a bug–the AI is a gap-filling machine, so consider:

1) which specific gaps did it fill

>this means you probably could stand to clarify this in your own synopsis and logline

2) precisely how it filled in these gaps

>does it come up with good ideas? Are they better than the ones you have? Are they something that might spark you to create a new, better idea of your own?

Remember, however, that the AI does tend to write very generalized loglines that often assume the reader is familiar with the screenplay already.* Use this as a tool for reflection: what is the AI assuming, and what would you like a reader to understand of your script from the logline/synopsis?

Make changes to your own version accordingly.

It hopefully goes without saying, but do not simply ask for a synopsis and use that. Use it as a tool to analyze your own work.

*Obviously this is a bit dumb, but we are where we are with AI. And it’s not anywhere near as far as the AI companies’ PR machine would have us believe.

Plugging the Holes in Your Story

The way I do this is to get the AI to read a 250-word synopsis, then ask it to:

1) paraphrase the synopsis

and

2) highlight any questions it has or gaps in the story.

Crucially, do not ask it to fill in these gaps. Remember, it is a gap-filling machine, so it is important to discourage it from doing so. In my experience, however, it will still attempt to fill gaps a bit–so be conscious of where it has added or conflated.

What you want to do here is to note where the AI’s version differs from your own. Note mistakes that it made as well as gaps that it acknowledged. Consider these places where you could be more specific.

Remember, don’t just take the AI’s prompt and run with it–after all, it’s almost certain to have added something you didn’t write! This is about learning to reflect on what the AI interprets and clarifying your own writing.

It’s worth repeating: just because the robot says it doesn’t make it right.

At later stages, you can try this same exercise with a 10-page outline, and perhaps even the entire script. Once you’ve fed these in, see how it gives you, for example:

–the story in 250 words, presentable to a prospective producer/financier

–a 25-word and 50-word logline, where the 25-word focuses on developing irony, conflict, and change within the story (a bare hook) and the 50-word introduces the main story engine (a hint toward how the hook is developed as a story)

–protagonist, antagonist, and supporting character bios

–use a common system such as the Save the Cat beats or the Hero’s Journey or John Yorke’s 5-Act Structure to break the script down

Remember, you may not agree with the robot’s assessment, but this is good. That means you’re thinking for yourself. You know, doing actual writing like a live human being.

Be on the lookout for any place where it gets weird–what happens when the AI doesn’t understand what you were saying? Did it hallucinate something to patch a hole? What is the hole? Is the hallucination worthwhile or just plain bonkers?

Which brings us to…

Do not use the AI prompt; rather, see what is correct or incorrect about the prompt and use this to clarify your own writing.

(I’m writing how to use AI as a screenwriter for SEO here.)

Using Hallucinations as a Goad

Any place that the AI isn’t reading your work properly…

1) means that it is either something is unclear in your writing;

2) the AI is hallucinating – however, if the AI is hallucinating, is this something useful that spurs your thinking about where to go next?

Use the AI’s answer as a prompt to make your own thinking clearer, more precise, and more creative

Remember, hallucinations are a feature, not a bug – AI thinks in terms of relationships rather than facts, and in a situation where relationships aren’t clear enough, it will invent facts to make the relationships work

Finally…

3) The Rushkoff Test

Media Theorist Douglas Rushkoff proposes that once you have the structure of a scene in mind, write your own version of the scene and then give the structure of the scene to the AI. Bear in mind that whatever the AI gives you is, of course, the statistically most likely thing to have happen.

One interesting way to look at this is that–given that it’s the most likely outcome–it is by definition uninteresting. Let the AI write the scene and assume that it’s giving you the most boring, pedestrian way to do things.

The rubber hits the road, of course, when the AI aligns with what you said. That doesn’t mean you’re “doing something right,” argues Rushkoff, but rather you’re just being pedestrian.

That means think about where your idea aligned with what the AI spit out–and then find something more interesting and divergent to say instead. That is, if you and the AI both got from A to C via point B, then this means you could almost certainly come up with a stranger, more creative version of point B.

Conclusion

Remember, AI is first and foremost a tool. A tool that is highly advanced, and extremely prone to error. That said, the error comes because it’s trying really hard to fill in gaps–and if you’re aware of this inclination, you can use it to your benefit.

Note where the gaps are–or where the AI hallucinated–and you’ll be able to see 1) how to clarify your story for producers; 2) figure out inconsistencies in your plot/story; and 3) find more creative ways to connect story points.

As long as you’re the one cracking the whip, AI can be a very useful tool for you as a screenwriter. AI is your tool; don’t be its tool.

Ignore this part

I’m just writing how to use AI as a screenwriter to appease the gods again