Why Transparency Matters
The reality is that many creators already use AI in their process, whether it's brainstorming ideas with a chatbot, generating character names, or rewriting a few sentences. But most of the time, audiences have no idea how AI was involved. That gap between what creators do and what readers know is where distrust grows.
Muse's AI usage system isn't designed to be “beaten” or “hacked.” Trying to evade accurate scoring defeats the purpose of the Muse platform. Muse provides a clear, honest way to show your audience how your creative process works — and lets them appreciate you for it.
When you're upfront about your process, you're building something more powerful than any single story: a fan base that trusts you. In the era of AI, honesty isn't a weakness. It's your biggest creative advantage.
Overview
Every Muse project tracks two independent AI usage scores, each on a 0–100% scale:
- Storyboard Score — how much AI contributed to your story's characters, settings, and plot.
- Manuscript Score — how much of the written text originated from AI versus your own typing.
0% means no AI was used. 100% means entirely AI-generated and unedited. The system gives generous credit for editing AI output, writing detailed prompts, and iterating on ideas.
Where Scores Appear
Scores are always visible to you as you write so there are never any surprises:
- Project Settings — charts show your storyboard and manuscript AI usage scores side by side.
- Exported PDFs & EPUBs — both scores are embedded in the document for verification.
- Verification Page — anyone with your document ID or exported PDF or EPUB can verify the scores.
The Color Scale
The score dial uses a continuous gradient. Lower AI usage is shown in purple; higher usage transitions to orange.
| Score Range | Color | Interpretation |
|---|---|---|
| 0 – 33% | Purple | Minimal to low AI involvement — mostly human-created |
| 34 – 66% | Purple → Rose | Moderate AI assistance with meaningful user direction and editing |
| 67 – 100% | Rose → Orange | Substantial to heavy AI generation with limited modification |
Transformation Score
Both the storyboard and manuscript AI usage scores rely on a shared transformation score that measures how much you've edited a piece of AI-generated content. It compares the original AI output to the current version, word-by-word, using a Longest Common Subsequence (LCS) algorithm.
Let Ts = the transformation score, A = the original AI words, and C = the current edited words:
The result ranges from 0 (identical, no edits) to 100 (completely rewritten). Rearranging sentences, adding words, and deleting AI text all increase the transformation score.
| Editing Amount | T_s | What it means |
|---|---|---|
| No changes | 0 | AI text left exactly as generated |
| Light edits | 10 – 25 | Fixed a few words or sentences |
| Moderate rewrite | 25 – 50 | Rewrote several sentences, kept the core ideas |
| Heavy rewrite | 50 – 75 | Significantly restructured; only fragments remain |
| Full rewrite | 75 – 100 | Almost entirely new text — very little in common with the original |
This score is referenced as Ts in the storyboard and manuscript formulas below.
Storyboard Score
The storyboard score measures AI contribution to your story's structure and foundational creative ideas. Any time AI modifies or creates a storyboard element, it is tracked and reflected in the score. The following actions count as AI contributions:
- AI editing — using AI features to rewrite or refine storyboard beats, characters, or settings
- Beat expansion — generating child beats from a parent beat using AI
- Gap resolution — using AI to fix plot holes detected by plot gap analysis
- Element generation — creating characters, settings, and full outlines with the AI brainstorming chat
If you build and edit your entire storyboard from scratch without using any of these features, the score is 0%. Stage 1 below applies whenever AI modifies elements. Stage 2 (engagement discounts) applies only when using the Ghostwriter brainstorming chat, which provides additional signals about how actively you directed the AI.
Stage 1: Element-Level Weighted Average
Every tracked element (character, setting, or beat) has an origin (user-created or AI-generated) and a transformation score Ts. User-created elements contribute 0. For AI-generated elements:
Each element carries a weight reflecting its importance in the story structure:
| Element Type | Weight |
|---|---|
| Root beat with children (section / chapter) | 10 |
| Character | 7 |
| Root beat without children | 5 |
| Non-root beat with children | 5 |
| Leaf beat (scene) | 3 |
| Setting | 3 |
The raw score before engagement usage discounts is:
Stage 2: Engagement Usage Discounts
Two discounts reduce the usage score based on how actively you directed the AI during the brainstorming process.
Phase 1 Discount — Characters & Settings
This discount considers how many words you typed in the story idea prompt. Let P = prompt word count:
| Prompt Length | Discount |
|---|---|
| 5 words or fewer | 0% |
| ~40 words | 2.4% |
| ~80 words | 5.2% |
| ~120 words | 7.9% |
| 150+ words | 10% (max) |
Phase 2 Discount — Outline / Beats
This discount combines three signals from your brainstorming chat session: the number of messages sent, the total words typed, and how much you edited the AI-generated storyboard plan.
Where Ts,plan is the transformation score of the AI-generated plan.
| Brainstorming Activity | Approximate Discount |
|---|---|
| 1 short message, no plan editing | ~1 – 3% |
| 3 messages, ~100 words, light plan edits | ~8 – 12% |
| 5 messages, ~250 words, moderate plan edits | ~20 – 25% |
| 6+ messages, 400+ words, heavy plan rewrite | ~35 – 40% (near max) |
Discounts are applied per-element: beat contributions are multiplied by (1 − D2), and character/setting contributions by (1 − D1).
Stage 3: Category Cap
If no outline beats exist yet (only characters/settings), the storyboard usage score is capped:
Final Storyboard Score
Estimated Score Table
| Score | Scenario |
|---|---|
| 0% | User created all characters, settings, and beats manually. No AI edits, expansions, or generation used. |
| 1 – 10% | User built the storyboard manually but used AI editing or beat expansion on a few elements, then heavily rewrote the AI output (60%+ rewrite). |
| 10 – 25% | AI edited or expanded several beats and characters. Significant user rewriting afterward. Or: Ghostwriter generation with detailed brainstorming and heavy editing. |
| 25 – 40% | AI edited or generated a substantial portion of the storyboard. User edited beats meaningfully (30 – 50% rewrite). |
| 40 – 60% | AI generated full storyboard, or AI editing heavily used for expansions and edits. User made moderate edits (20 – 40% rewrite) to some elements. |
| 60 – 80% | AI generated full storyboard. User made light edits (10 – 25% rewrite) — tweaking names, adjusting descriptions, minor restructuring. |
| 80 – 95% | AI generated everything with minimal prompting. User accepted most elements with only minor tweaks (<10% rewrite). |
| 96 – 100% | AI generated all elements. User accepted everything as-is with zero or near-zero edits. |
Manuscript Score
Muse computes the manuscript AI usage in two stages: a word-level count of AI versus user text, and a per-operation prompt direction usage discount.
Stage 1: Text Origin Score
Muse's editor invisibly tracks every span of AI-generated text. The algorithm counts words inside AI spans versus words you typed directly. For each AI text span, a power-curve formula reduces the AI weight based on the transformation score Ts:
The power curve drops faster than a straight line, giving you more credit for edits.
| Amount Edited | T_s | AI Weight (A) |
|---|---|---|
| None | 0 | 1.00 — counts fully as AI |
| Light (a few words) | 10 | 0.85 |
| Moderate | 25 | 0.65 |
| Significant | 50 | 0.35 |
| Heavy rewrite | 75 | 0.13 |
| Complete rewrite | 100 | 0.05 (near zero) |
Stage 2: Prompt Direction Discount
When you write a detailed instruction to direct an AI operation (rewrite, expand, edit), the AI weight for that text is slightly reduced. Inline completions and full drafts (no user prompt) receive no discount. Let P = your prompt word count:
| Prompt Length | Discount on AI Weight |
|---|---|
| No prompt (inline completion / full draft) | 0% |
| ~20 words | 2% |
| ~40 words | 4% |
| ~60 words | 6% |
| 80+ words | 8% (max) |
Final Manuscript Score
Sum up effective AI words and divide by total words:
Estimated Score Table
| Score | Scenario |
|---|---|
| 0% | User wrote everything manually. No AI text inserted. |
| 1 – 5% | User accepted a handful of inline completions (1 – 2 sentences each) across a full manuscript. All other text typed by hand. |
| 5 – 15% | User wrote most of the manuscript. Occasionally used rewrite/expand tools on a few paragraphs with detailed prompts. |
| 15 – 30% | User wrote the majority but used AI writing tools (rewrite, expand, tone shift) on multiple sections. Moderate editing of AI output afterward. |
| 30 – 50% | Significant AI assistance. Several sections drafted by AI then heavily edited (50%+ rewrite). Or many writing tool uses across the manuscript with moderate edits. |
| 50 – 70% | AI generated a substantial portion. Multiple full-draft sections with moderate editing (25 – 50% rewrite). Writing tools used frequently. |
| 70 – 85% | Most sections AI-drafted. User made light edits (10 – 25% rewrite) — fixing sentences, adjusting dialogue, minor tweaks. |
| 85 – 95% | Nearly all sections AI-drafted via Ghostwriter level 5. User accepted drafts with minimal changes (<10% rewrite). Some inline completions accepted as-is. |
| 96 – 100% | Full manuscript generated by AI. User accepted all drafts with zero or near-zero edits. No manual writing. |
AI Dial Levels & Scoring
Muse's AI Dial lets you choose how much AI assistance you want, from Level 0 (no AI) to Level 5 (full generation). Not every level introduces features that affect your AI usage scores. Here's how each level maps to scoring:
| Level | Name | Scoring Features | Affects Score? |
|---|---|---|---|
| 0 | Solo | None — pure writing environment | No |
| 1 | Spellcheck+ | Spelling and grammar only — no AI-generated text | No |
| 2 | Nudge | Creative prompts appear when you pause — informational only, never inserted into your work | No |
| 3 | Copilot | Inline sentence completions (manuscript) and idea suggestions that can be applied to your storyboard | Yes |
| 4 | Collaborator | AI editing, beat expansion, gap resolution, Ask Collaborator — all tracked as AI contributions | Yes |
| 5 | Ghostwriter | Full storyboard generation, full manuscript drafts, brainstorming chat — all tracked | Yes |
Levels 0 – 2: No Impact on Scores
At levels 0 through 2, no AI-generated content is ever inserted into your storyboard or manuscript. Spellcheck corrections and creative nudge prompts are purely informational and are never tracked as AI contributions. Your scores will remain 0% at these levels.
Level 3 (Copilot): Light AI Contributions
Level 3 introduces two features that can affect your scores:
- Manuscript — inline sentence completions. When you accept a suggestion, the inserted text is marked as AI-origin and counted toward the manuscript score. Suggestions you dismiss have no effect.
- Storyboard — idea suggestions for beats, characters, and settings. If you apply a suggestion, the created or modified elements are tracked as AI-edited and counted toward the storyboard score. Viewing a suggestion without applying it has no effect.
Because Copilot contributions tend to be small (a sentence here, a single element there), scores at this level are typically very low — often under 5%.
Level 4 (Collaborator): Directed AI Editing
Level 4 unlocks AI tools that edit and expand your existing work based on your instructions. Every operation is tracked and reflected in the scores:
- Storyboard — Ask Collaborator to edit beats, characters, and settings via instruction. Expand beats into sub-beats. Detect and resolve plot gaps. Each modified or created element is tracked as an AI contribution.
- Manuscript — Ask Collaborator to rewrite, expand, compress, or adjust the tone of selected text. The edited text is marked as AI-origin. Inline completions from Level 3 remain available.
At this level, you are always in control — AI only acts on specific selections or elements that you point it at. The transformation score gives you credit for editing AI output afterward.
Level 5 (Ghostwriter): Full AI Generation
Level 5 is a significant step up. It unlocks the Ghostwriter — an interactive brainstorming chat that can generate entire storyboards and manuscript drafts from scratch. This fundamentally changes AI's role from an editor to a co-creator:
- Storyboard — the Ghostwriter chat can generate full character rosters, settings, and multi-level beat outlines in one session. All generated elements are tracked. The engagement discounts (prompt length, message count, plan editing) apply here to reward active brainstorming.
- Manuscript — Ask Ghostwriter can generate full section drafts from your storyboard beats, producing entire scenes or chapters at once. This typically results in higher manuscript scores than Collaborator-level edits, since more of the raw text originates from AI.
All Level 4 features remain available at Level 5. The key difference is the scale of generation — Collaborator edits what you've written, while Ghostwriter can create from a blank page.
Document Verification
When you export a manuscript as a PDF or EPUB, Muse calculates both scores at export time and embeds them in the document. Each export receives a unique document ID (e.g., MUSE-A1B2C3-D4E5).
Anyone — a reader, agent, or publisher — can visit the Verify page, enter the document ID or upload the PDF or EPUB, and see:
- The storyboard and manuscript AI scores
- The AI levels that were active (e.g., Level 4 — Collaborator)
- The project name, author, word count, and export date
Scores are frozen at export and cannot be changed after the fact. This system is designed to build trust between writers and their audiences.
A Note on External AI Detection
Muse doesn't only track AI usage within the platform. We also detect when content has been generated or substantially written using external AI tools outside the platform.
When external AI usage is detected, the affected usage scores are automatically set to 100% for the given project to reflect the true nature of the content. We won't go into the specifics of how this works — but we're confident in our approach and are continuously improving it.
To avoid having your AI usage score set to 100%, we strongly recommend that you write all project content yourself (with or without Muse's AI tools) inside the Muse platform.
This isn't about catching people or punishing anyone. It's about preserving the integrity of what Muse is built for: honest creative attribution. Readers, agents, and publishers who verify a Muse document deserve to know what they're looking at — and you deserve credit for the work that's genuinely yours.
The best way to use Muse is simply to be honest. Write with AI, write without it, or do anything in between — Muse is designed to reflect your process accurately, whatever that looks like. That transparency is what makes your work trustworthy, and that trust is what builds lasting relationships with your audience.