What a Script Report Can Describe, and What It Cannot Know
The pages can be described. The film, the actor, and the buyer cannot, because they are not in the file.
A coverage report is a reading of a file. Some of that reading can be checked against the file tonight. Some of it is a guess about a film that does not exist yet. Mixing those two is how a useful note becomes a false certainty. No report was commissioned for this article.
What you can check against the draft
Scene notes, a repeated beat, a character who never chooses, a speech that explains a fact the audience already has. If the report names a page, open the page. Quilty’s report guide says a full feature report includes scene notes, a coverage grid, and a revision roadmap. ScriptReader.ai says paid tiers include scene-by-scene notes. Either way, the check is the same: is the thing on the page?
Quilty’s limitations page says models can invent citations. A paragraph about page 64 is a claim until you see page 64. You got your screenplay coverage is the practical sort, and the worksheet is there so the check has columns.
A logline is also checkable. It either describes the draft or it describes a different film. What the report thinks your screenplay is is that test.
What the file does not contain
The actor’s performance. The director’s cut. The buyer’s slate. Yesterday’s location quote. Whether an audience will forgive the ending after they have sat in a theater. The report can estimate some of these, and Quilty’s guide says it estimates audience, casting strategy, and a budget. Estimates are not the events.
The limitations page is plain about the mechanism. Models synthesize plausible text. They are not a database of your production. Provenance is often opaque. Outputs change when the model changes. A sentence that sounds like insider knowledge of “the market” may be a pattern, not a phone call.
Quilty’s methodology page uses The Blair Witch Project to show a miss: an era-locked 1999 score of 58, and a statement that the model would not have predicted the outcome, because the comparable did not exist yet. That is the shape of “cannot know.” The film had not taught the category. No report written before the fact contained the audience that showed up.
How to label the sentences
On the printout, mark each paragraph “in the draft” or “about the film.” Rewrite from the first pile. Keep the second pile in a memo until a person who has the missing fact can use it. A producer can use a budget guess as a question. They should not use it as the budget. Budget notes are not a cut list.
Can AI script coverage replace the first read? is the companion point about taste and the decision to send the script out. This piece is narrower: even a careful first read cannot contain facts that are not in the world yet.
Takeaways
- Check every cited scene against the draft before you rewrite toward it.
- Audience, casting, and budget sections are models. Label them as models.
- A confident miss, including Quilty’s Blair Witch example, is a reason to keep the comps visible.
- The film is not the PDF. Do not grade the PDF as if the film had already been shot.
Sources checked October 8, 2026
- Quilty report guide. Scene notes, roadmap, audience, casting strategy, and production plan.
- Known AI limitations. Hallucinations, provenance, and unstable outputs.
- Quilty Score methodology. The Blair Witch example.
- Quilty help. Casting Lab.
- ScriptReader.ai. Scene-by-scene notes on paid tiers.