eBook Doctor

App privacy policy

This is the complete privacy policy for eBook Doctor, the app for ChatGPT operated by Global Scholar Press (a trading name of Nobtrex LLC, United States) at https://globalscholarpress.com/mcp. It states every category of data the app receives, what each tool sends back to ChatGPT, why we process it, who else sees it, how long we keep it, and how you delete it.

Contact: gsp@walzone.com · Support · Terms · Privacy policy of the website and the book-writing service

1. What the app is

eBook Doctor is a set of 52 tools that ChatGPT calls on our server to inspect, repair, validate, convert and prepare EPUB files, and to help you review the writing chapter by chapter. Our server runs no AI model of its own: it executes exactly the operations ChatGPT requests, using deterministic programs (our EPUB engine, the official EPUBCheck validator and Calibre for format conversion). It is free and needs no account.

This policy covers the app endpoint /mcp, the download endpoint /dl/… and the book pages /book/…. The rest of globalscholarpress.com — the marketing site, the free online EPUB editor and the paid book-writing service — is covered by the website privacy policy.

2. Categories of data we collect

2.1 The files you attach

The EPUB you attach in ChatGPT, and any cover image or replacement file you ask ChatGPT to put into the book. ChatGPT passes our server a temporary download link and we fetch the file from it; files up to 80 MB are accepted. Two fallbacks exist for the same purpose: you can give a public URL (our server then makes an ordinary web request to that address, so the site hosting it sees our server’s request, not your device) or small files can be passed inline as base64. Your original upload is stored untouched and is never modified.

2.2 The arguments of the tool calls

ChatGPT sends the arguments each tool needs. Across the whole app these are:

• the workspace identifier (book_id) of the book being worked on;
• paths of files inside your book, resource kinds, line ranges and chapter numbers;
• text from your book and text you want in it — search terms and regular expressions, replacement text, the exact old and new text of a patch, whole replacement resources, and paragraph-addressed chapter edits;
• book metadata you or ChatGPT drafted — title, subtitle, series, author names, publisher, language, rights, subjects, dates, description or blurb, 7 store keywords, 3 BISAC category codes, e-book and paperback prices, page count, trim size, paperback and e-book ISBNs, target audience, the copyright statement and the answer to the store’s AI-disclosure question;
• chapter-review notes — a status, a short note and an issue count you or ChatGPT record for a chapter;
• operation options — checkpoint labels and ids, conversion target format, severity filters, and the trim size, page count and paper type for the cover-geometry calculation.

Nothing else is requested. The tools ask for no profile data, no “just in case” fields and no conversation context beyond the arguments of the call.

2.3 Identifiers we create

When a book is opened we generate a random 32-character workspace identifier, and when files are delivered a random book-page token. They are not derived from you or your device, they identify one book, and they exist only as long as the workspace does.

2.4 Technical request data

Our web server and application record, for each request, the time, the endpoint or tool name, the HTTP status, the size of the response, the browser identification string and the network address the request came from. Tool calls arrive from OpenAI’s servers, so for the app that address is OpenAI’s infrastructure, not your device. If ChatGPT sends an MCP session header we log only its first 12 characters; at the time of writing ChatGPT sends none. Opening your book page in your own browser is an ordinary web request, so that one does carry your address.

2.5 What we never receive

We do not receive your OpenAI account, your name, e-mail address or phone number, your chat history, or anything you type in ChatGPT that is not an argument of a tool call. We create no account, set no cookie and store no identifier in your browser from the app. We ask for no location, no payment details, no government identifiers, no credentials and no health data. Please do not place such data in a tool call; if you do, it is deleted with the workspace.

3. What the tools send back to ChatGPT

Every tool result goes to ChatGPT, which shows or summarises it for you; OpenAI then processes it under its own privacy policy. Results are limited to what answers the call. A result can contain:

• the workspace identifier, the file name you uploaded, its size and EPUB version;
• the book’s own metadata read from the file — title, author(s), publisher, language, identifier or ISBN, date, rights, subjects, description and cover details — and the list of files inside the book with their kinds, media types and sizes;
• content of your book when you ask for it: the text of a resource or chapter, numbered paragraphs, search hits with surrounding context, the package and navigation files, and differences between versions;
• validation output: EPUBCheck codes, severities, messages and the file, line and column they refer to, the version of EPUBCheck used, our plain-language diagnosis, the before/after error counts of this session and the time of each check, the KDP readiness checklist and the store rules it quotes;
• checkpoint ids, labels, kinds and creation times, and what changed between them;
• the manifest of the listing-metadata package you saved and the names of the files in it;
• the book-page link, how many hours it remains valid, and short instructions addressed to ChatGPT about what to do or report next;
• a small links block with three addresses on globalscholarpress.com — the free EPUB editor (carrying your workspace identifier so it opens this book), our home page and the book-writing service page. They carry a campaign tag, utm_source=chatgpt, that tells us a visit came from the app. It identifies the app, not you, and a tool result sets no cookie and no advertising identifier.

The complete field-by-field list of what a result can contain is in section 4. It is enforced by our server: every result is checked against that list and any field not on it is removed before it is sent.

Results never contain: our server’s file paths or directory names, program or converter logs, process exit codes, stack traces, the network address the call came from, internal diagnostic, debug or telemetry fields (such as byte offsets, internal issue ids or editor status flags), or anything belonging to another person’s book. If you want ChatGPT to see less of your book, ask it to read only the files or chapters you name.

4. Tool-by-tool data map

All 52 tools, grouped by what they do. “Receives” lists the arguments in addition to the workspace identifier.

Every field a tool result can contain (337)

This is the complete list. Before a result leaves our server, every field that is not on it is removed, so a result can hold only these fields, nested exactly as named (a.b is field b inside a; a[] is each item of list a; a.* is a count keyed by severity or kind). There are no hidden, debug or diagnostic fields. Most fields appear only in the results of the tools that need them.

Workspace and the uploaded file

Book metadata read from your file

Saved store-listing package (save_metadata / get_metadata)

Facts handed to ChatGPT to draft a listing (suggest_metadata)

Files and content of your book

Chapters and writing review

Edits and their effect

Checkpoints and differences

Validation

Amazon KDP readiness and cover

Delivery and conversion

Instructions and links

5. Why we process this, and on what basis

• To perform the operation you asked ChatGPT for. This is the only use of your files and their content. In the EU/EEA, the UK and Switzerland the legal basis is performance of the service you requested.
• To keep the service working and safe. Request logs let us find errors and block abuse. Legal basis: our legitimate interest in a functioning, secure service.
• To count usage in aggregate. We count tool calls per day and visits that arrive from the app. These counts contain no book content and no per-person profile. Legal basis: legitimate interest.

We never use your book or its content to train AI models, never sell or rent data, never use app data for advertising or profiling, and never make automated decisions about you. The app displays no advertising.

6. Who else sees the data

• OpenAI — ChatGPT is the channel: it sends us your file and the tool calls and receives every result. What OpenAI keeps is governed by OpenAI’s privacy policy and your ChatGPT settings.
• IONOS SE — the hosting provider of the server in the United States on which the workspaces sit. It acts on our instructions and has no right to use the contents.
• Google (Google Fonts) — only if you open your book page in a browser: the page loads a web font from Google’s font service, so your browser’s network address reaches Google for that request. No book data is sent.
• No one else. The advertising-measurement tags used on our marketing pages (described in the website policy) are deliberately absent from the app endpoint, the download endpoint, the book pages and this policy page. No advertising or analytics company receives app data. We would disclose data only if legally compelled, and we hold nothing that identifies you.

7. Where the data lives, and for how long

Each book gets its own private workspace on our server, readable only by our application, addressed only by its random identifier; workspaces are never listed and cannot be browsed. Everything produced for a book — the repaired EPUB, converted formats, the metadata package and the reports — is delivered from one book page, globalscholarpress.com/book/<random token>, which is marked noindex and carries no advertising code. Anyone who has that link can download those files, so treat the link like the files themselves.

8. Your choices and rights

• Delete everything now: ask ChatGPT to close the book, or press Erase my book from this server on your book page. Deletion is immediate and permanent on our side. Otherwise it happens automatically within 24 hours of the last activity.
• Share less: tell ChatGPT to read only the chapters or files you name — the tools only return what they are asked for.
• Access, correction, deletion, objection, portability: write to gsp@walzone.com. We reply within two business days and act within 30 days. Because a workspace is tied to no identity, we can act on one only while you can give us its book-page link or identifier.
• EU/EEA, UK, Switzerland: the legal bases are in section 5; you may complain to your data-protection authority. California: we do not sell or share personal information and app data is not used for cross-context behavioural advertising; you may exercise your CCPA rights by e-mail.
• Children: the app is not directed to children under 13 and we do not knowingly collect their data.

9. Security

All connections use HTTPS. Workspaces are stored with owner-only permissions, are isolated by unguessable identifiers, are never listed, and no login exists that could expose them. Our staff open a workspace only to investigate a problem you report to us.

10. Changes to this policy

When this policy changes we update the date below, and material changes are announced on the support page. The version of this policy in force is the one published here.

Published 22 September 2026; updated 9 October 2026 (complete field-by-field list of tool results, enforced by the server). Operator: Nobtrex LLC (Global Scholar Press), United States · gsp@walzone.com