Five clear settings
Choose 0%, 5%, 15%, 25%, or 40% to match your review policy.
Human ideas. Helpful tools. Your standards.
Bring more context to AI detection with an allowance that fits the way you write.
A single score does not tell the whole story. Choose your threshold, then review the result in context.
Find your allowanceSelect an allowance
15% allowance · the raw score stays at 20%.
One writing workspace, with your review policy in view. Explore the allowance settings in this interactive illustration.
Your raw score stays unchangedA first draft might begin with a person’s research, include an AI-assisted outline, and end with careful human editing. Other documents may be written entirely without AI. These workflows call for clear expectations, not assumptions.
AI Allowance gives you a way to apply your own review threshold to GramProof’s AI score. It helps you distinguish the detector’s estimate from the amount of assistance your writing policy permits.
Choose 0%, 5%, 15%, 25%, or 40% to match your review policy.
Keep the detector’s estimate separate from your interpretation threshold.
Discuss permitted assistance and document expectations before reviewing a result.
Use allowance alongside the existing scan results and available sentence highlights.
Think of allowance as a review boundary. If the score is at or below the selected value, it sits within your allowance. If it is higher, it sits above your allowance and deserves a closer look.
For example, a score of 20% is above a 15% allowance and within a 25% allowance. Switching settings changes that comparison; it never turns the raw score into a different number.
Allowance is not a word-count quota. A 25% setting does not establish that one quarter of a document may be generated, or prove that the remaining text is human-written.
| Review approach | Classic AI Detection | AI Allowance |
|---|---|---|
| Decision boundary | A fixed threshold for interpreting a score. | A choice of five interpretation thresholds. |
| Writing policy | The reviewer must apply context separately. | The chosen allowance makes the review boundary visible. |
| Raw detector score | An estimate from the configured detector. | The same estimate; allowance does not change it. |
| Human review | Still necessary to understand a result. | Still necessary; no threshold proves authorship. |
Choose a threshold that reflects an agreed policy. These settings guide interpretation; they are not measurements of the proportion of words written by AI.
| Allowance | Review style | When to consider it |
|---|---|---|
| 0% | Very strict | Review every AI-like signal against a policy that permits no AI assistance. |
| 5% | Strict | Use a small interpretation margin for a tightly controlled writing workflow. |
| 15% | Moderate | Set a middle-ground review threshold when limited assistance is acceptable. |
| 25% | Flexible | Use a more permissive threshold for a workflow that allows AI assistance. |
| 40% | Very flexible | Choose a higher review threshold for explicitly AI-assisted drafting. |
Your voice, your process
Clarity before submission
Context for every review
Shared client expectations
Consistent writing policies
| Who it is for | How allowance can help | What to review alongside it |
|---|---|---|
| Writers | Align a draft with a publication’s AI-assistance policy. | Research notes, drafts, and editing decisions. |
| Students | Understand the limits set for a specific assignment. | Course rules, citations, and disclosed tool use. |
| Teachers | Use one consistent review boundary for a defined task. | Student discussions, prior work, and revision history. |
| Agencies | Agree on client expectations before drafting begins. | The brief, editorial review, and contributor records. |
| Schools | Explain shared expectations across a writing program. | Institutional policy and a fair human review process. |
A “99%+ accuracy” headline is meaningful only with a defined dataset, method, language, model version, and independent evaluation.
GramProof has not supplied a validated benchmark supporting that figure here. Allowance changes your review boundary—not the detector’s accuracy.
Use context, not a percentage alone.Paste or upload a complete passage in the AI Checker. Use at least 100 words to provide context.
Select the threshold that fits your assignment, publication, or organization’s agreed policy.
Use Scan for AI Score and let the configured detector analyze the submitted text.
Compare the score with your allowance, then review available highlights, confidence, and writing history.
AI Allowance makes your expectations easier to express, but a thoughtful review still begins with the writing and the person behind it. Set a clear policy, discuss permitted tools, and use GramProof to support that conversation.
False positives and false negatives are possible. Never use a detector score as the sole evidence of authorship or misconduct.
It is the threshold you choose when interpreting an AI score. GramProof lets you select 0%, 5%, 15%, 25%, or 40% to fit the rules of your writing workflow.
Build a writing review workflow with clear expectations and room for human context.
Create your GramProof account Already have an account? Sign in to continue to your dashboard.