Manuscript AI

The workflow is deliberately short. You tell Manuscript AI a little about your book — including any specific questions you want answered — then upload the manuscript as a .docx file. The system runs the manuscript through its own AWS-hosted models and, where useful, through paid APIs from OpenAI, Anthropic and Google, and returns a structured report.

That report has four main parts. A structural analysis answers eight editorial questions covering the book's main message, writing style, flow between chapters, accessibility and whether readers will know what to do next. A chapter-by-chapter section gives word counts, estimated reading times, Flesch reading ease and Flesch–Kincaid grade level scores, a summary and concrete recommendations. A dedicated section responds to the author's own questions, and a recommendations block prioritises what to change first.

Manuscript AI is explicit about data handling. Connections are TLS-encrypted, storage disks are encrypted, and the uploaded file is deleted automatically within 48 hours of the report being delivered (or within 48 hours of upload if you never check out). Manusus are excluded from AI training under the terms of the third-party APIs used, the company trains no models of its own, and reports remain at a private link until you request deletion. Payment goes through PCI DSS-compliant gateways and the service claims no rights over your manuscript.