How to Write a Software Method Patent: AI Auto-Generates the 'Method-Apparatus-Device-Medium' Four-Part Suite
CNIPA.AI Team
Tech Blog
In Chinese invention patent applications, software method patents (IPC G06) are the undisputed number one category, accounting for over 32%. However, software patents present unique drafting challenges: how do you 'translate' an algorithm or process into a technical solution that meets patent law requirements? Based on our data analysis of 31,210 recently published patents, this article details the standard software patent drafting patterns and how to use AI tools to efficiently generate compliant patent documents.
What Is the Software Patent 'Four-Part Suite'?
Our data analysis shows that 29 patents (5.8%) include 'storage medium' in their titles, while 169 patents (33.8%) contain both 'method' and 'system/apparatus'. Behind these numbers is China's standard software method patent drafting paradigm — the 'Four-Part Suite': The first part is the method claim, describing algorithm execution steps (e.g., 'A data processing method comprising: S1 acquiring input data; S2 processing through a neural network model...'); the second is the apparatus claim, mapping method steps to functional modules (e.g., 'A data processing apparatus comprising: a data acquisition module, a model processing module...'); the third is the device claim, describing a generic hardware architecture of processor + memory + program instructions; and the fourth is the storage medium claim, protecting the computer-readable storage medium carrying the program instructions. The core logic is: using a single technical solution to build comprehensive protection from method, apparatus, device, and medium dimensions. This is the 'gold standard' for Chinese software patents, recognized by both examiners and patent agents.
Data-Driven Writing Guide for Titles and Claims
Based on statistical analysis of 31,210 patents, we've distilled the following writing parameters: For titles — average 19.5 characters, median 20; 62.2% start with '一种' (a type of...), 8.4% with '基于' (based on...); standard format is 'A [innovation]-based [purpose] method, apparatus, device, and storage medium'. For claims — G-class (computer) patents average 9.9 claims with a median of 10; recommended 2-3 independent claims (covering method, apparatus, medium) plus 7-8 dependent claims. Dependent claims expand into: specifying step details, defining parameter ranges, adding preferred implementations. For drawings — G-class patent median is 14 (average 57, skewed by patents with numerous UI screenshots), with core drawing types including system architecture diagrams, method flowcharts, module block diagrams, data flow diagrams, and sequence diagrams. Mastering these data-driven parameters makes patent drafting significantly more standardized and efficient.
How AI Auto-Generates the Software Patent Four-Part Suite
CNIPA.AI's AI drafting feature has been specifically optimized for software method patents. Simply input a technical description (500+ words recommended), and AI automatically completes: Step 1 — Technical element extraction: AI identifies core algorithm steps, input/output data, and key technical features; Step 2 — Method claim generation: generates the main method claim in S1/S2/S3 step format and automatically expands 3-5 dependent claims; Step 3 — Four-part suite auto-mapping: automatically converts method claims into apparatus, device, and medium claim sets, ensuring one-to-one correspondence of technical features; Step 4 — Specification generation: includes all five required sections (technical field, background, invention content, drawing descriptions, detailed embodiments). In testing, AI-generated software patents scored above 85% across claims completeness, specification adequacy, and format compliance. For corporate applicants, AI drafts can serve directly as internal review versions; for patent agencies, AI drafts reduce drafting time from 8-16 hours to 2-3 hours of review and polishing.
Get Started with CNIPA.AI
Sign up now and experience AI-powered patent search and writing
Sign Up Free