What Screenwriters Actually Do Before Page One (2026)

Key Takeaways

• Screenwriting software starts at page one. Almost everything that determines whether a script works happens before that, in tools that were not built for it.
• Five artefacts exist before the first slugline: the want, the ending, the shape, the world file, and the scene list. Most first drafts fail because two or three of them were skipped.
• The most common skip is the ending, and it is the most expensive, because a scene's job is only definable once you know what it is building toward.
• Almost nobody does this in Final Draft or Highland, and they are right not to: page-one software is for the draft, not the development, and forcing development into it produces an outline pretending to be a script.
• The stage is spatial before it is linear. Index cards persist in screenwriting for the same reason they persist in editing: comparison requires seeing everything at once.

Page One Is the Middle of the Process

Ask a screenwriter what they use and you get Final Draft, Highland, Fade In, WriterDuet, or Scrivener. All of those are page-one tools: they exist to produce correctly formatted pages, and they are good at it.

Then ask what they were doing for the six weeks before they opened one, and the answer is a notebook, a wall, a Notion page, a pile of index cards, a document called `thoughts.txt`, and a lot of walking.

That gap is not a tooling failure so much as an unexamined one. The work that determines whether a script functions almost entirely precedes page one, and it is done in improvised systems because the software everyone names starts too late.

Five artefacts exist in that space. They are not a process to follow in order, and a first draft that fails usually turns out to be missing two or three of them.

The Five Artefacts

One: the want

What the protagonist is trying to get, stated as an action rather than a feeling.

"She wants to be respected" is a feeling and it generates no scenes, because nothing can be attempted. "She wants her name on the paper she wrote" is an action, and it generates scenes immediately: things she can try, people who can refuse her, ways she can escalate.

The test: can you write down three things she does to get it, and can somebody stop her doing them? If not, the want is not yet a want, and this is the single most common root cause of a script where nothing seems to happen despite events occurring.

Two: the ending

How it finishes, concretely: the last image, what has changed, and what the audience knows now that they did not.

This is the most commonly deferred artefact and the most expensive to defer, because a scene's job is only definable relative to what it is building toward. Without an ending, you cannot know whether scene eleven is doing work or filling space, so the middle collapses and the writer blames the middle.

The ending is allowed to change. Almost every one does. What matters is that one exists, because a provisional ending organises the writing and no ending does not.

Three: the shape

The five to nine turns, not a full beat sheet. The points where the direction changes: the thing that forces the story to start, the reversal, the point of no return, the lowest moment, the choice at the end.

Whatever model you use, or none, the shape is where pacing lives. Turns clustered together produce a rushed section; a long gap between them is the flat middle everybody suspects and cannot locate.

Keep this loose. A full forty-beat sheet before drafting is where scripts go to become transcription, and transcription is flat, because the discoveries that make a scene alive happen while writing it.

Four: the world file

Everything you need consistent: names, ages, timeline, geography, who knows what and when, the rules if there are rules.

This one is unglamorous and it prevents the specific humiliation of the continuity error found by a reader in draft three. It is also where research lives for anything with a real subject.

The rule that matters: the world file is a reference, not a creative document. Time spent expanding it is usually time spent avoiding the script, and worldbuilding is the most respectable procrastination available to a writer.

Five: the scene list

One line per scene: who is in it, where, and what changes by the end.

That last clause is the whole value. A scene where nothing changes is a scene where nothing happens however good the dialogue is, and writing the line forces the admission. A scene list is also the first artefact that can be reordered cheaply, which is where structural problems become visible.

Fifty to seventy lines for a feature, eight to fifteen for a short.

Why It Is Spatial

The reason index cards have survived every generation of screenwriting software is not nostalgia. It is that four of the five artefacts require comparison, and comparison requires seeing everything at once.

You cannot tell that scenes fourteen and twenty-two do the same work by reading them in order, because reading is sequential and the observation is spatial. You cannot see that the protagonist stops making choices for six consecutive scenes when you can only see two on screen. You cannot notice that a subplot vanishes for a third of the middle by scrolling a list.

A document holds the sequence. A wall holds the structure. They are different views of the same material and only one of them shows you the problems.

This is why the standard setup persists: cards or a wall for development, then a page-one tool for the draft, then back to the wall when the draft has a problem nobody can locate.

What Screenwriters Actually Use

Two honest observations from this table.

Scrivener is the tool that spans the most of it, because its corkboard sits directly on the manuscript, so a card is a scene rather than a picture of one. For writers who dislike running four tools, it is the strongest single answer at around $59 once.

Nothing in the page-one category serves the development stage, and that is a reasonable scope rather than a flaw. Trying to develop in Final Draft produces an outline formatted like a script, which reads as progress and is not.

For the spatial half, a canvas that persists is useful because development happens over months with gaps, and a wall gets taken down. Storyflow, which we make, holds the shape, the scene list, the world file and the research on one board the AI reads, so questions like which scenes do the same work, where the protagonist stops choosing, and which threads go quiet longest are answerable from the material rather than by rereading. It is not a screenwriting tool and the draft belongs in Final Draft or Highland. It is paid-only during early access from $7.99 a month annual, with the Free plan landing before the end of 2026, and index cards on a table remain excellent and cost almost nothing.

Where AI Helps and Where It Does Not

On the want, as an interrogator. Give it your protagonist's want and ask what three actions it generates and who can block them. If it struggles, the want is a feeling.

On the ending, adversarially. Describe your ending and ask it to argue that nothing has changed. If that case is easy to make, you have a mood rather than an ending.

On the scene list, diagnostically. Reading fifty scene lines and reporting which do the same work, where the protagonist stops making choices, and which characters vanish is comparison across the whole set, which is slow by hand and exactly what the wall is for. This only works if the model can see all of them at once.

On the world file, as search. Answering "how old is she in the second act" from your own notes.

Not on the pages. Generated screenplay dialogue is competent and dead, and it is competent in a way that is recognisable to anybody who reads scripts professionally, which is the audience.

Not on the shape. A generated structure is the median structure for your genre, which is what you were presumably trying to escape.

What Goes Wrong

Starting at page one. The script gets written from an idea rather than from a want and an ending, and it collapses somewhere in the middle for reasons that feel mysterious and are structural.

No ending. Every scene becomes undecidable, and the middle takes the blame.

A full beat sheet before drafting. Writing becomes transcription and the prose goes flat.

Worldbuilding as avoidance. The most respectable procrastination in writing, and the world file grows while the script does not.

A want that is a feeling. Nothing can be attempted, so nothing can be blocked, so there is no story despite plenty of events.

Developing in a document. The sequence is visible and the structure is not, so scenes doing duplicate work stay invisible until a reader points them out.

No "what changes" on the scene list. Without it a scene where nothing happens looks identical to one where something does.

Never rebuilding the wall after draft one. The diagnostic is most valuable on an existing draft, and almost nobody goes back to it, which is when it would repay most.

The Bottom Line

Screenwriting has excellent software that begins at exactly the wrong moment. Final Draft, Highland, and Fade In are all good at producing formatted pages, and by the time you need them, the decisions that determine whether the script works have already been made or skipped.

Five artefacts live before page one. What the protagonist is trying to get, stated as something they can attempt. How it ends. The five to nine turns. The reference file for whatever must stay consistent. And a scene list where every line says what changes.

The two most commonly skipped are the ending and the want, and skipping either produces the same symptom: a script full of events where nothing seems to be happening, and a middle that collapses for reasons that feel mysterious and are entirely structural.

None of that work is linear, which is why index cards outlived every piece of software that was supposed to replace them. A document shows you the order. A wall shows you the shape, and the shape is where the problems are.

So do the five things somewhere you can see all of them at once. Then open the page-one software, which is genuinely excellent at the job it was built for, and which was never built for this one.