How readability is measured
We combine formula-based signals with structural checks. Formula outputs estimate reading difficulty from measurable language features; structural checks identify the sentences and patterns that may slow comprehension. Results are presented as editing evidence, not as a guarantee that a document is correct, persuasive, or suitable for every audience.
What formulas are used
- Flesch Reading Ease: estimates ease from sentence length and syllables, with higher values generally indicating easier text.
- Flesch-Kincaid Grade Level: expresses similar signals as an approximate United States school-grade level.
- Gunning Fog: emphasizes sentence length and complex-word density to highlight dense business or editorial prose.
- SMOG: uses polysyllabic words as a strong signal and is useful for comparing formal or public-information content.
Formulas may disagree because they weigh language features differently. We recommend reading results as ranges, looking for agreement across signals, and inspecting the underlying sentences before making an editorial decision.
How issue detection works
Detection highlights long sentences, difficult words, passive voice risk, and hard-to-scan structure patterns. Suggestions are prioritized so teams can fix high-friction lines first.
A detected issue is a review prompt, not an automatic error. Passive voice can be appropriate when the actor is unknown or intentionally secondary. Long sentences can remain readable when their relationships are clear. The editor decides whether a flagged pattern harms the intended reader in context.
Where automated rewriting helps
Task-specific rewriting accelerates simplification and plain-English adaptation, especially for support content, onboarding copy, and dense policy summaries.
The model receives a task-specific instruction rather than a generic request to improve the text. Output is then mapped into a predictable result structure with a summary, issues, suggestions, revised text, and next actions. Users should still compare every material claim with the original.
Limitations
Model outputs can improve readability but may require context checks for legal precision, regulated language, or internal policy nuances.
- Formulas do not measure factual accuracy, logical completeness, or legal effect.
- Automated issue detection can produce false positives and miss domain-specific ambiguity.
- Automated rewrites may omit qualifications or choose terminology that is inappropriate in context.
- Grade and CEFR outputs are practical targets, not certified assessments of a reader or document.
When manual review is required
Use manual review for legal commitments, policy exceptions, compliance wording, or any text where wording changes could alter enforceable meaning.
Manual review is also required for medical guidance, safety instructions, financial claims, technical procedures, accessibility-critical content, and public communication where an omission could cause harm. The reviewer should verify facts, numbers, defined terms, conditions, warnings, links, and the action a reader is expected to take.
Recommended flow: start in Readability Checker, refine in Text Simplifier, then validate with domain review.