The $20 First-Pass: A Notes-Then-Polish Workflow Before You Pay a Human Reader

Published by StoryNotes. We sell the $20 AI first-pass this workflow is built around, so the conflict of interest runs straight through the piece. Every price and every failure case is sourced so you can check the math without trusting us.

Most writers buy coverage the way people buy gym equipment: one big purchase, at the wrong time, that's supposed to fix everything. The coverage market prices reads from $9.99 to $400+, and almost all the advice about it is a which-service argument. The more useful question is a when question. The same $169 spent in the right order buys three drafts of feedback; spent in the wrong order it buys one expensive diagnosis of a script you already knew was broken.

TL;DR: Sequence reads by draft stage. Free passes while the story is still finding itself. A $20 AI structural pass on each working draft — cheap enough to repeat, consistent enough to measure the rewrite. One human read ($129–$195) when the story is locked and taste is the open question. A three-draft cycle runs about $169 sequenced, against $387+ for a human read on every draft. Each step has an exit criterion; the workflow is knowing them.

Because a read's value depends on what the draft can do with it — and drafts change what they need. A first draft has structural problems: acts that don't turn, a protagonist whose goal blurs at the midpoint, a subplot that vanishes for thirty pages. A polish draft has taste problems: voice, tone, whether the thing is alive. Those are different diagnoses, and the market prices them differently — structure is cheap now ($9.99–$29 from AI tools), taste is not ($100–$195 from a professional).

Buying out of order fails in both directions. A $129–$195 human read on a first draft pays a professional for hours of judgment on scenes that won't survive the rewrite. And an AI pass on a polish draft grades structure you've already fixed while staying silent on the one question left — everything downstream of taste, which is exactly where the documented AI failures live.

There's also a reliability reason to hold the expensive read until the end. Human scores on early drafts are noisy: in one unusually well-documented case, screenwriter Arnon Shorr's identical draft drew Black List scores of 8, 7, 6, and 5 from four evaluators (arnonshorr.com, 2021). All four flagged the same flaw; they weighed it differently. A single early human score is one opinion buying a lot of authority.

One concrete fix: Before you spend anything, write down which draft you're on — honestly. "Story still moving" and "story locked" are the only two categories the workflow cares about.

Exhaust the free tier — but know what it's for. Free feedback in 2026 means peer swaps, self-audits, and DIY chatbot passes with ChatGPT or Claude. Prompted well, a chatbot is genuinely useful at scene level: dialogue that repeats a beat, a scene entered too early and left too late, description bloat. What free passes reliably miss is full-script structure — token limits and generic prompts produce notes on the pages the tool actually held in view, not the shape of the whole thing.

The free tier is also where you use your cheapest reader: yourself, after a two-week drawer. No purchase in this workflow outperforms rereading your own script cold.

One reassurance before any AI touches your pages: this is not a rules problem. The WGA's 2023 MBA provisions on AI govern employment terms on covered projects — nothing in them restricts what tools you point at your own spec (WGA).

Exit criterion: you've fixed everything scene-level you know how to find, and the open questions are structural — does act two work, does the protagonist's arc track, do the setups pay off. That's a full-script question, and it's where the free tier stops.

One concrete fix: Give the free pass a written goal before you run it ("find every scene that ends without a turn"), and when your goals start containing the word "act," move to step 2.

A structural map of the whole script, in minutes, cheap enough to be repeatable. This is the AI tier — $9.99 (ScriptReader.ai) to $29 (Prescene), with StoryNotes at $20 — and this is the step we sell, so here is the claim at its honest size. In the Editors Guild's 2025 adversarial test, run by the union representing Hollywood's own story analysts, AI held its own on loglines and lost to human readers "hands down" on notes and analysis (Variety, 2025). Independent practitioner tests document the same ceiling: hallucinated details, uniform praise for weak scripts, intentional structure misread as error (Act Four Screenplays, 2025).

What survives that evidence is exactly the job this step needs: structure, setup-payoff tracking, and scene-level pacing are where our own five-tool test found AI notes most reliable. A $20 pass on a draft you already know needs a rebuild is a different economic object from a $129 read you wait ten days for — you're buying a map, not a verdict. What a StoryNotes pass returns: a full-script report in about five minutes, letter grades, page citations in the format "p.48 · INT. KITCHEN," and two to three concrete rewrites per section (sample report here — judge the output, not the pitch).

Two rules keep this step honest. First, never buy AI coverage for a verdict — the Editors Guild test caught AI handing a sycophantic recommend to a script that circulated Hollywood for twenty years without selling. If a tool sells you a PASS/CONSIDER/RECOMMEND line, ignore that line. Second, don't chase notes you disagree with; AI notes are hypotheses with page numbers, and the documented failure modes — hallucinated details, intentional choices misread as errors — mean some of them will be confidently wrong.

Exit criterion: you have a prioritized structural punch list you believe, ordered by what breaks the most downstream — and you haven't fixed any of it yet. That's the signal to stop reading notes and start rewriting.

One concrete fix: Run the pass, then sort every note into three buckets — agree, disagree, can't tell. Rewrite from the agree bucket only. The can't-tell bucket is your question list for the human read in step 4.

Because consistency is the AI tier's one superpower, and the delta is the product. The same rubric applied to draft two and draft three actually measures whether the rewrite worked — the thing four human readers demonstrably don't do the same way. Rereading the new report against the old one answers the only question that matters mid-rewrite: did the act-two fix hold, or did it move the problem thirty pages downstream?

This is also where the economics of the tier show up. A second $20 pass brings the running total to $40 — still less than a third of one entry-level human read — and keeps the feedback cadence inside a rewrite loop instead of outside it: notes tonight, rewrite this week, notes again next week. Standard human turnarounds of 10–14 days fit a polish schedule fine; they fit an active restructure badly.

Exit criterion: two consecutive passes stop surfacing new structural problems, and the grades on the sections you rewrote moved. When the report starts repeating itself, the machine has told you everything it knows how to see. More AI passes past that point are reassurance, not information — stop buying them.

One concrete fix: Read the delta before the notes: put the two reports side by side and check the sections you rewrote first. If a rewritten section's notes didn't change, the rewrite didn't land — that section goes back on the punch list.

Exactly once, on the locked draft, when taste is the open question. A professional reader — Coverage Ink runs $129 standard, 10–14 pages with a 36-category grid; Script Reader Pro assigns named screenwriters at $169; Industrial Scripts runs $195+ with vetted consultants (all verified July 2026) — spends hours with your script and renders what no rubric holds: whether the voice is alive, where the script sits against what's selling, and the note about absence — the scene the structure is begging for that doesn't exist. Software grades what's on the page; the absence note is the human superpower, and it's worth the whole fee.

The workflow's earlier steps exist to protect this purchase. By the time the human reads it, the structural problems are gone — so the taste judgment lands on the story you meant, and the can't-tell questions from step 2 have a professional answer. If you will only ever pay for one read of a finished script, pay a human; that sentence costs us money and it's true.

Two buying rules. Pick a service that names its readers or matches genre — you're buying judgment, so buy a known judge. And never act on one score alone: the 8/7/6/5 variance case is the argument for either budgeting two reads on the draft that matters or treating a single evaluation as one professional's opinion, because that's what the data says it is.

Exit criterion for the whole workflow: the human read comes back with taste-level notes instead of structural ones. Fix those, and the next dollar goes to querying, not coverage.

One concrete fix: Send the human reader your logline and your step-2 can't-tell list with the script. You're paying for hours of professional attention — aim them at the questions the machines couldn't answer.

$169 for the core three-draft cycle at July 2026 prices, with well-marked exits above and below. Here is the whole ladder:

The comparison bar is the argument. The sequenced path and the human-only path buy feedback on the same three drafts; the difference is that the sequenced path spends taste money on taste questions and structure money on structure questions. The $218 you don't spend is two more human reads' worth of diagnosis on scenes that were scheduled for demolition.

Budget-flexing the ladder is straightforward. Broke: substitute a rigorous free DIY pass at step 2 and accept the full-script blind spots. Flush: add the Black List evaluation at querying time — $100 plus $30/month hosting, budgeted ideally as two evaluations, because of everything the variance case taught us.

One concrete fix: Write the whole sequence down with dollar amounts before buying anything. If your plan has a human read before the story is locked, you've found the line item to cut.

When the calendar or the goal changes what a read is for. Three legitimate exceptions:

• A deadline owns the schedule. Contest and fellowship windows don't wait for exit criteria. If the draft ships Friday, a fast structural pass now beats the right read too late — and Coverage Ink's $49 five-day semi-rush exists for exactly this squeeze.
• The score is the product. If you're buying a Black List evaluation for discoverability — the score feeds a real industry pipeline — you're not buying notes, and step-4 logic applies early: go in with the story locked, or you're paying $100 to document problems you knew about.
• You already have a professional attached. A manager, a producer, or an assignment means notes arrive from the people who matter. Coverage of any kind becomes a private pre-flight check, not the main feedback channel.

What doesn't justify breaking it: impatience dressed as readiness. The workflow's whole function is to make "is this draft done?" an evidence question — the delta stopped moving, the human read came back about taste — instead of a mood. Writers who skip to the expensive read early don't get a better script; they get a professionally formatted description of the rebuild they were avoiding.

Start where the workflow starts: run the free pass this week. When it stops being enough, the $20 first-pass is here — and the sample report is public, so you can judge the output before spending anything.

One concrete fix: Put the exit criteria in your writing calendar as actual gates — "no human read until two AI passes agree the structure holds" is a rule you can follow on a bad day, which is when the $200 mistake usually happens.