Step 2: Prepare reference materials

Important note

Please note that AI writing on this platform does not search for and piece together mixed-quality information from the internet; it relies entirely on the reference materials you provide.

Therefore, reference materials are the starting point of the creation process—the more reference materials you provide and the higher their quality, the better the quality of the resulting book and the lower the duplication rate.

We recommend preparing reference materials related to the writing topic in advance (usually a few to a dozen e-books related to the writing topic).

We have over 500TB of materials across various categories, which can support writing in most fields. If you lack materials, you can contact the platform staff.

Supported formats

In this system, the file formats that the AI writing feature can read and use include:

• md (Markdown)
• TXT
• EPUB
• PDF
• DOCX (Word)

Among these:

• PDF files should be double-layer files (that is, they should have a text layer). For single-layer PDF files that only have an image layer, you can use the PDF image content extraction feature in the toolkit to first extract and convert the content into a text-format file;
• Word files usually have two extensions, DOC and DOCX, where DOC is an earlier file format. Please make sure to save your Word files as DOCX format.

How the amount of material affects AI writing

The system requires each writing project to select at least 2 reference books before entering AI analysis. If fewer than 2 are selected, Save Selection and Resource Validation cannot proceed, and the prompt “Please select at least 2 reference books” is displayed.

In addition to the minimum quantity, the system also provides a Capacity Match prompt based on the ratio of “total word count of reference materials” to “total target word count of the project” (displayed after you click Resource Validation):

Recommendation: The larger the target length (for example, more than 300,000 words for the whole book), the more and thicker reference books you should prepare; if you have only one or two thin volumes, you should appropriately lower the expected length (total word count) in the project.

The impact of material quality on AI writing

In addition to word count scale, the topic relevance and file usability of the materials are equally critical:

• Topic relevance: Before analysis begins, the system uses AI to score each material against the current book topic (0–100). Materials scoring below 45 are marked as “Low relevance” and a prompt warns that this may affect final book quality. You can choose to remove irrelevant materials or continue using all materials.
• Main file format: A resource can be used for book-writing analysis only when its main file is one of md / TXT / EPUB / PDF. Although formats such as Word and PPT can be stored in the Resource Library, they cannot be selected in a writing project.
• Resource status: Disabled resources do not appear in the selectable list; resources without a main file also cannot be used for analysis.
• Resource metadata: The title, author, description, and other details are used in relevance evaluation; accurate entries help the system determine whether a resource is off-topic.

The Resource Library supports reusing the same reference book across multiple writing projects, but each project is parsed, analyzed, and generated independently without affecting the others.

Associating resources with projects

You first need to associate certain resources with a project. During AI writing, these resources will be listed first for you to select.

How to do it: In the resource list, select some resources and click "Associate with writing project" in the right-click menu

Use this platform as cloud storage

Even if you do not use the platform's AI book-writing features, you can use this project's resource management features as cloud storage to upload, reliably store, manage, and use various files:

• Upload and archive: Use New Resource, Bulk Upload, or drag and drop to the list area to back up local files to the cloud.
• Categorization and search: Search by title or author; filter by resource type, enabled status, pinned status, writing project, and more, making it easy to quickly locate files among many.
• Online preview: PDF, Markdown, EPUB, images, audio and video, etc. (subject to format support) can be previewed in the browser without downloading each time.
• Download and distribution: download a single original file, or select multiple items and then Download selected files to retrieve them locally in batches.
• Permissions and isolation: Files are stored in the space of the current tenant and isolated from other tenants on the platform; operations such as create, edit, and delete are controlled by role permissions.
• Capacity visibility: You can view cloud storage usage in the console on the Home page (vendor-hosted or tenant-owned OSS, subject to actual configuration).

File types supported by Resource Management

In Resource Management, in addition to the 5 file formats dedicated to book writing, you can upload files in various formats and types. These files will not be used as reference materials for writing, but can still be saved on this platform (equivalent to using this platform as your cloud drive):