🛒 AI novel assistance solution for online literature authors and content creators, covering worldview construction, outline planning, chapter writing, style polishing, AI continuation, multi-language translation and publishing, and IP cross-form incubation.

AI novel assisted creation and IP incubation program

Solution overview

Track background

The scale of China's online literature market will exceed 40 billion yuan in 2025, and the scale of the overseas translation market will simultaneously grow to the tens of billions level. Platforms such as Qidian.com, Tomato Novel, and QQ Reading have added tens of thousands of chapters on average every day, and competition is fierce. Authors face a hard threshold of "updating 4,000 to 10,000 words per day," and the continuous output of high-quality content has become a core pain point. At the same time, large language models represented by ChatGPT and Claude have made breakthrough progress in Chinese long text generation, style imitation, and narrative structure control. AI-assisted writing has moved from "able to write" to "usable, controllable, and deliverable".

The current penetration rate of AI writing among online article authors has exceeded 35%, but most of them are stuck in the superficial stage of "let AI write a paragraph and try it", lacking a systematic full-process workflow design. This program aims to fill this gap - not just a list of tools, but a set of executable methodologies from conception to publication, from single chapter to full text, from text to IP.

Target user portrait

• Online article contracted authors: Contracted writers who have serialized works on Qidian, Tomato, Qimao and other platforms need to maintain a daily update of 4,000-10,000 words
• Studio Writer Team: A creative studio with content mass production as its core, where multiple people collaborate to complete the operation of series works
• Published Writer: For authors with Amazon KDP or physical publishing, complete manuscript writing and multi-language translation capabilities are required
• IP incubation operator: a content operator who transforms novels into comics, short plays, and audiobooks
• Writing Novice: Beginner users who want to try novel writing and need structured guidance.

ROI expectations

• Daily update speed: 3000-5000 words/day purely manually → 8000-15000 words/day after AI assistance (increased by 2-3 times)
• Complete this cycle: A 300,000-word novel shortened from 6-8 months to 2-3 months
• Polishing rework rate: Reduce post-revision workload by about 60%
• Cost structure: The average monthly investment in AI tools by individual authors is about 100-300 yuan, and that of studios is about 500-2000 yuan.
• Translation overseas: The translation cycle of a 200,000-word Traditional Chinese/English version is shortened from 4 weeks to 3-5 days, and the cost is reduced by 80%

Preconditions

• Have a network environment with normal access to major LLM platforms
• Open the paid version of at least one mainstream AI conversation tool (ChatGPT/Claude/DeepSeek/Kimi/Doubao, etc.)
• Understand the upload rules and format requirements of the target novel platform
• Have basic word/typing and document management skills

Toolchain list

Tools	Purpose	Required Account Level	Estimated Fees	Alternatives
ChatGPT	Outline generation, chapter writing, polishing, continuation	Plus/Pro ($20-200/month)	Starting from $20/month	Claude
Claude	Long context consistency check, worldview construction, deep polishing	Pro ($20/month)	$20/month	DeepSeek
DeepSeek	Stylized Chinese web writing, low-cost batch writing	API pay-as-you-go	~¥0.5-2/10,000 words	Kimi
Kimi	Long document analysis, chapter summary, plot logic review	Free/Pro	Free/¥30/month	豆包
豆包	Domestic network optimization, daily Chinese writing	Free	Free	Tongyi Qianwen
Wawa Writing	One-stop writing workbench for web articles, including outline/chapter/continuation	Free/paid	Free/¥30-100/month	Sudowrite
Grammarly	English writing proofreading, grammar polishing	Free/Premium	Free/$12/month	Platform built-in proofreading
Midjourney	Character illustrations, cover designs, scene concept drawings	Paid ($10-60/month)	Starting from $10/month
ElevenLabs	AI audiobook dubbing, character voice customization	Paid ($5-22/month)	Starting from $5/month	Platform TTS tools
Total			Individual approximately ¥150-800/month

Note: The above fees are reference prices in mid-2026. The actual prices are based on the latest pricing page of each tool. It is recommended that novices start with the free version + Kimi/Doubao combination, and upgrade to the paid plan after stable output is confirmed.

Preparation

• [ ] Register LLM account: Open at least 1-2 paid accounts of mainstream AI tools (ChatGPT Plus + Claude Pro + DeepSeek API recommended)
• [ ] Install writing tools: Register a Wawa Writing or Sudowrite account and configure the writing workbench
• [ ] Create material library: Create a local or cloud disk folder system (material/outline/chapter text/setting set/polished draft)
• [ ] Open a publishing platform account: Register for Qidian.com Author Area/Tomato Novel Creator Center/Amazon KDP Account
• [ ] Prepare prompt word library: Collect or build a set of system prompt words for novel scenes (world view generation/character setting/writing style imitation/continuation guidance)
• [ ] Set writing goals: clarify the target word count, daily update plan, completion node and publication path
• [ ] Test tool availability: Confirm that all accounts can log in normally and API calls are normal.

Step-by-step guide

Step 1: World view and character construction

⏱ Estimated time: 2-4 days 🎯 Goal: Produce a complete world view setting document and core character biographies, and establish a consistent "creation Bible" for subsequent writing ⚠️ Prerequisites: Novel type (fantasy/urban/romance/suspense, etc.) and core creative direction have been selected

Operation instructions

This is the most critical "laying the foundation" stage in the entire process. If the world view is not solid, there will inevitably be inconsistencies in the settings in subsequent writing; the character arcs are not clear, and readers cannot establish emotional empathy. The task of AI at this stage is not to imagine for you, but to use structured questions to help you turn vague inspiration into executable setting documents.

Specific operation path

1. Basic world view settings (using Claude or ChatGPT)

Use the following prompt to guide AI to generate a worldview framework:

You are a senior web worldview architect. I need to build a worldview for a [writing genre, such as Eastern fantasy/urban supernatural/cyberpunk] novel. Please analyze according to the following framework and fill in one by one:

1. Basic rules of the world: power system/level of technology/source of magic (examples)
2. Geographic pattern: key regions, power distribution, civilization level
3. Timeline: chronology of core historical events (at least 10 key nodes)
4. Social structure: class division, power structure, economic system
5. Race/species settings: no less than 3 main races, including differences in appearance, culture, and abilities
6. Source of conflict: the core conflict that drives the plot (at least 3 levels)

Only one module is output at a time, and I will continue to the next one after confirmation.

2. Generate character biographies (using ChatGPT or Kimi)

Generate complete character profiles for each main character:

Creating a core character profile for my novel. Role positioning: [protagonist/villain/female lead/second male lead/mentor, etc.].
Please include:
-Basic information: name, age, appearance characteristics
- Personality portrait: MBTI type, core personality tags (3-5), mantra/habitual actions
-Backstory: 3 turning events in my growth experience
- Ability system: skill tree/training level/special talent
- Motivation-driven: short-term goals, long-term desires, and what the bottom line cannot touch
- Character arc: the change path from the starting state to the ending state of the story
- Relationship network: a map of interpersonal relationships with other core figures

3. Power system/level setting (use DeepSeek for deep digging)

DeepSeek's generation of Chinese fantasy terms is of high quality:

Design a power hierarchy system for Eastern fantasy novels. Requirements:
- At least 9 levels (from low to high, each level is named with a Chinese aesthetic style)
- Each level describes the qualitative change after breakthrough (dimensions such as body/soul/domain/lifespan)
- The same level is divided into four stages: early stage/middle stage/late stage/peak stage.
- Give 2-3 reasonable explanations of the mechanism of "leapfrog combat"
- Refer to the common routines of Tomato novels/Starting Point but with innovative differences

4. Set up document summary

Summarize all the content output by AI into a "Work Setup" document (it is recommended to use Feishu Document/Notion or local Markdown) as the only reference source for subsequent writing.

Verification method

• Setting documents can answer questions such as "What are the rules of this world?" "Why did the protagonist make this choice?"
• There are at least 3 core characters who can interact naturally "without AI"
• The character relationship diagram is logically self-consistent: there is no shortcoming of "the two don't know each other but the plot requires their cooperation"

FAQ

• Q: What should I do if the AI-generated worldview is too universal? A: Each time the AI ​​is required to provide 3 differentiated options, choose the most anti-routine one and then deepen it. Question: "Where is the originality of this setting among the works on Qidian.com?"
• Q: Are the character biographies and subsequent writing not consistent? A: Add the character biography to the AI ​​system prompt before writing, and do a character consistency scan every 5 chapters.

Step 2: Outline and Chapter Planning

⏱ Estimated time: 3-5 days 🎯 Goal: Output the outline of the book, volume planning, detailed outline of the first 30 chapters, and clarify the distribution of exciting points/foreshadowing/climax ⚠️ Prerequisites: World view setting document and character setting completed

Operation instructions

The biggest difference between online writing and publishing writing is "rhythm design" - when to upgrade the protagonist, when to bury suspense, and when to explode conflicts, which directly determines the reader retention rate. The core value of AI at this stage is to turn "it feels like it should be written like this" into "structured planning supported by data."

Specific operation path

1. Main line outline of the book (use Claude to handle long context)

Based on the following world view and character settings, plan a novel outline for me with an [estimated word count] of 10,000 words.
Structural requirements:
1. The book is divided into 3-4 volumes, each volume is marked with the core conflict theme
2. List 5-8 key plot points in each volume (plot points, conflict escalation, reversals)
3. Mark "distribution of cool points": the design idea of at least 3 climax/explosion scenes in each volume
4. Mark "Foreshadowing Management": Where to recycle the foreshadowing buried in the previous volume?
5. Mark "Rhythm Curve": the beginning, transition and conclusion of each volume (description of narrative rhythm within 500 words)

2. Planning by volume and chapter by chapter (using ChatGPT’s thinking chain capability)

For the first volume "[Volume Name]", please generate a chapter-by-chapter outline of the first 30 chapters. Format:
Chapter X:
- Core event (one sentence)
-The emotion/message this chapter wants to convey to the reader
- Appearing characters
- Foreshadowing (if any)
- The hook at the end of the chapter (suspense/anticipation)
- Cool design (if there is one in this chapter)
- Estimated word count: 2000-3000 words

3. AI-assisted proofreading outline logic (using Kimi’s long document analysis capabilities)

Paste the complete outline document to Kimi and ask:

Please analyze this outline from the perspective of a "senior online article editor":
1. Do the first 3 chapters grab the reader? Suggested areas for improvement
2. Are the plot connections between volumes natural? Are there any gaps?
3. Is the main character’s growth line clear and coherent?
4. Is the current foreshadowing setting too obvious or too subtle?
5. Overall rhythm: Is there a risk of "slow at the front and fast at the end" or "water injection in the middle"?
6. From the perspective of platform editors, what are the chances of approval and the selling point of this outline?

Verification method

• The outline can answer "What to write in each of the first 30 chapters?"
• The outline can answer "Is there a clear conflict → upgrade → resolution path in the main line of the book?"
• The outlines of the first 3 chapters are detailed enough that you can jump right into writing

FAQ

• Q: What should I do if AI’s outline is too routine? A: Add "Avoid the following routines: breaking off the engagement/counterattack by losers/picking up the old man/substituting people at a time-travel wedding" in the Prompt. Ask the AI to provide "a way of handling this plot line that is different from 80% of similar works"
• Q: Will writing too detailed an outline cause you to lose the joy of writing? A: The detailed outline can be written down to the level of plot nodes. The specific dialogue, scene description, and emotional rendering can be left to the creation of the main text.

Step 3: Text section creation

⏱ Estimated time : Daily duration (recommended 2-4 hours for each writing session) 🎯 Goal: Stably output chapters at a daily update rate of 4000-8000 words, maintaining consistency in writing style and narrative quality ⚠️ Prerequisite: Chapter outline completed and confirmed

Operation instructions

Text creation is the part with the largest workload. The core role of AI is "enlargement engine" and "style maintainer" - based on the plot nodes in the outline, AI generates a first draft or draft paragraphs, on which the author modifies, optimizes and adds truly touching details.

Specific operation path

Mode A: Detailed outline→AI generated first draft→Manual polishing (recommended for novices)

Enter a single chapter outline into ChatGPT Plus or Claude:

Please create a chapter of the novel based on the following outline. The word count is 2500-3000 words.
[Detailed outline content pasted]
Writing requirements:
- Writing style reference: [Provide a 200-word personal essay]
- Use third-person limited perspective (follow the protagonist's perspective)
- Dialogue accounts for 30-40%, including action description and micro-expressions
- There is a small hook (suspense/reversal/golden sentence) every 500 words
- Avoid writing routine expressions, such as "a flash of surprise flashed in his eyes"
- The end of this chapter needs to create a suspense of "I want to read the next chapter"

Mode B: AI-assisted writing (suitable for experienced writers)

Use API tools such as DeepSeek to complete the expansion of the first draft, and the author focuses on refinement:

Expand the following outline into a chapter. Style requirements:
- Use [author’s name]’s narrative rhythm (paragraphs 3-5 sentences long, dialogue short and frequent)
- The description of the scene should be limited to 15% of the full text, and the dialogue should promote the plot.
- At least 1 "highlight moment" in each chapter (emotional explosion/battle climax/reversal reveal)

Mode C: AI synchronous writing + manual review (to adapt to the pressure of daily updates)

The detailed outline of 5 chapters per week is handed over to AI at once, and a first draft of 1 chapter is generated every day, and is released after manual revision is completed on the same day.

Configuration parameter suggestions

• ChatGPT: GPT-5 or GPT-5.1 model, temperature 0.7-0.8 (creative type), Top P 0.9
• Claude: Claude Opus 4.x or Sonnet 4.x, temperature 0.6-0.7 (better narrative consistency)
• DeepSeek: DeepSeek-V4, temperature 0.8 (Chinese web text native support is better)
• Batch Prompt: Use system message to fix writing style settings, and only enter the detailed outline changes for each chapter

Verification method

• The reading experience of a single chapter is smooth, without obvious AI accent (such as "however", "but", "it is worth noting" and other transition words)
• After reading 3 consecutive chapters, the characters and writing style are consistent, and there is no gap between them.
• After modification, the "human touch" of the AI-generated content accounts for no less than 60%
• After completing the daily update, do a self-examination of "reader immersion": whether each chapter has the motivation to turn the page

FAQ

• Q: What should I do if the content written by AI is too "AI flavor"? A: Add a sentence at the end of the prompt "Please rewrite if the following situations occur: two consecutive sentences begin with 'him'; words such as 'however/however/it is worth noting' appear; more than 3 sentences of description do not advance the plot." You can also use Grammarly or a self-built prompt to remove the AI flavor after generation.
• Q: Does it require a lot of modifications but is it slower than pure handwriting? A: It means that the detailed outline is not detailed enough or the prompt lacks stylistic anchoring. It is recommended to invest more time in the detailed stage first and treat AI as an "expander" rather than a "creator"
• Q: Does the dialogue written by AI read like a question and answer? A: Added constraints to the dialogue prompt: "Each dialogue turn should not exceed 3 sentences; each dialogue contains at least one action description or psychological activity; avoid two characters saying 'um', 'okay' and 'yes' to each other."

Step 4: Style polishing and consistency check

⏱ Estimated time: half a day per week 🎯 Goal: Ensure a unified writing style throughout the book, no logical loopholes, and character OOC detection ⚠️ Prerequisite: Completed the text creation of more than 5 chapters

Operation instructions

This is a link that many AI writers tend to overlook. When AI generates long-form content, as the context window changes, problems such as "the writing is exquisite in the early stage but turns into vernacular in the later stage" and "the character is very calm in the first 20 chapters, but the character deviates after 50 chapters". Regular consistency checks can avoid major overhauls after completion.

Specific operation path

1. Style consistency check (using Claude’s long context advantage)

Input the completed 10 chapters of text + initial setting style sample text into Claude:

Please analyze the following content from the perspective of a "professional copy editor":
1. Is the writing style consistent across these 10 chapters? Specifically point out where the style has deviated.
2. List 3 overused words in the text and give replacement suggestions
3. Is the dialogue style consistent with the initial setting of the character? If there are OOC (out of character) issues, please specify the chapter and line number.
4. Is the density of scene description reasonable? Which chapters have too much description and slow down the pace, and which chapters have too little description and lack the sense of picture?
5. Give 3 targeted polishing suggestions

2. Logical vulnerability detection (using Kimi’s long document analysis capabilities)

As a novel logic checking expert, please scan the following content:
- Are there any contradictions in the timeline? (The character was at location A one day and appeared thousands of miles away the next day)
- Is the strength system consistent? (The protagonist was defeated in Chapter 5 and suddenly became invincible in Chapter 10)
- Are there any breaks in the relationships between the characters? (It was said before that the two were enemies, but later they suddenly formed an alliance)
- Was the foreshadowing forgotten? (The artifact mentioned in Chapter 3 has not yet appeared in Chapter 50)
- Are the names/titles consistent? Don't appear "Li Xuan called Brother Xuan in Chapter 3 and Xiao Li in Chapter 10"

3. Grammarly/GPT polishing pipeline

Pass the text of each chapter in order:

• AI detection of overused words (customized Prompt analysis with ChatGPT)
• AI de-smell processing (eliminating obvious traces of AI generation)
• Grammarly English novel part grammar check (if there are English paragraphs)
• Manual final reading (the most critical step, cannot be skipped)

Verification method

• Randomly select 3 chapters, and self-assessment of writing style consistency reaches over 85%
• The logical vulnerability detection report is "no obvious flaw"
• No OOC (Out of Character) issues

FAQ

• Q: What should I do if big problems appear every time I check the consistency? A: This is not a bad thing. It is recommended to do a consistency check every 10 chapters and add repair suggestions to the system prompts in subsequent chapters to prevent the problem from spreading.
• Q: After AI polishing, the "roughness" of the original work was lost? A: When polishing, distinguish between "flaw repair" and "style modification" - the former must be corrected (linguistic defects/logic/inconsistency), while the latter can be modified or not (rhetoric/rhythm/style preference)

Step 5: Multi-platform publishing and typesetting

⏱ Estimated time: about 1 hour for each release 🎯 Goal: Complete the release of domestic platforms such as Qidian/Tomato/QQ Reading + Amazon KDP English version preparation ⚠️ Prerequisite: Complete at least 30% of the manuscript of the book

Operation instructions

Different platforms have different layout rules, chapter word count preferences, sensitive word filtering and recommendation mechanisms. Pasting directly may cause layout confusion, sensitive words being blocked, or chapters being rejected. The value of AI in this link is format conversion, sensitive word pre-checking and multi-version adaptation.

Specific operation path

1. Platform format preprocessing

Use AI to convert the text into the format required by the target platform:

Please convert the following novel chapters into Qidian Chinese website publishing format:
1. Each paragraph should not exceed 200 words, with one blank line between paragraphs.
2. Start the dialogue on a new line and use Chinese full-width quotation marks.
3. Remove all Markdown tags
4. Check and replace sensitive words: [Customized platform sensitive word list]
5. Add a "Chapter End" mark at the end of the chapter
6. Count the actual number of words in this chapter (excluding spaces and punctuation)

2. Tomato novel/Qimao platform adaptation

Tomato novels have high requirements on rhythm and paragraph length:

• No more than 3 lines per paragraph
• At least 3 "hooks" (suspense points) in each chapter
• The first 200 words must grab the reader
• AI checks the "bounce rate risk" of each chapter and gives suggestions for improvement

3. Amazon KDP typesetting preparation

Prepare English/Traditional Chinese version for publication:

Please convert the following Chinese novel chapters into KDP-compatible format:
1. Convert to standard EPUB/MOBI compatible Markdown
2. Use Heading 1 for chapter titles
3. The first line of a paragraph is indented by 2 characters, and the paragraph spacing is 0.5 lines.
4. Generate directory tree
5. Add copyright page template
6. Check the compatibility of Chinese characters and special characters

Verification method

• Check the layout in the platform preview mode before publishing and there are no abnormalities.
• Sensitive word self-test passed, no blocked chapters
• KDP files are verified by Amazon for e-book format

FAQ

• Q: What should I do if I fail the Qidian review? A: First use AI to pre-check sensitive words ("Please list the words and plots in the above content that may be blocked by Qidian"), and then modify them one by one before submitting.
• Q: What should you pay attention to when publishing simultaneously on multiple platforms? A: The launch time difference between each platform is at least 24 hours to avoid violating the exclusivity clause. After the VIP chapter is first released on Qidian, updates on other platforms will be delayed for 3-7 days.

Step 6: Data-based operation and feedback optimization

⏱ Estimated time spent: 2-3 hours per week 🎯 Goal: Optimize chapter rhythm, follow-up rate and reader stickiness through data drive ⚠️ Prerequisite: At least 20 chapters have been published

Operation instructions

Internet articles are a "data-driven" content industry - read rate, completion rate, collection/recommendation ratio, and readers' sentiments directly affect recommendation traffic. Authors who write purely by hand often only focus on "whether the writing is finished" and ignore "whether the readers like the plot in this direction." AI can help collect, analyze and give improvement suggestions.

Specific operation path

1. Follow-up rate analysis

Analysis tools: Use Qidian/Tomato’s writer backend data and cooperate with AI analysis
Import the data of the last 10 chapters (number of chapters, number of readers, chapter retention rate):
Please analyze:
1. What do the three chapters with the lowest reading rates have in common? (The number of words is too short/the amount of information is insufficient/the plot is bland)
2. What do the three most read chapters have in common? (With reversal/emotional explosion/upgrade)
3. Based on the data trends, what adjustments are recommended in the next three chapters?
4. The impact of the current update frequency on pursuit reading: is it better to maintain data more frequently or less frequently?

2. Reader Comments AI Summary

Feed the last 50 reader comments (positive + negative + neutral) into the AI:
Please summarize:
1. TOP5 positive keywords most frequently mentioned by readers
2. Top 5 negative feedbacks most frequently mentioned by readers
3. What are the suggestions that "can be changed" (pacing/chapter length/character creation)
4. What are the directions that "cannot be changed" (complete genre positioning, core plot direction)
5. Is there a design that "some readers are dissatisfied with but overall it is in the right direction"
6. Suggest adjustments to the writing strategy in the next stage (if any)

3. Chapter AB test

The same chapter provides two different styles of beginnings (for example: scene setting style vs. direct conflict style), and different trumpets are posted to observe: -Which one has a higher reading rate for the first 200 words?

• AI analyzes the differences in narrative styles between the two
• Solidify better writing methods into the generated prompts in subsequent chapters

Verification method

• The follow-up rate has been increasing or stable for 3 consecutive weeks
• Negative comments account for no more than 15% of the total comments
• Complete data review at least once a week and adjust writing strategies

FAQ

• Q: What should I do if the reading rate data is not made public? A: The Tomato Novel Creator Center provides relatively complete reader data, and Qidian also has backend data after signing the contract. In the early stage, you can use the collection/recommendation ratio as an approximate indicator.
• Q: Will comment analysis overly cater to readers? A: Distinguish between "suggestions related to writing quality" (pace/length/typos) and "suggestions related to plot direction" - the former is actively adopted, while the latter remains author-led but can be used as a reference signal

Expected results

Specific output

Output	Specifications	Completion time
Work setting collection	5000-10000 words world view + character document	Week 1
Full outline + 30 chapter details	including rhythm curve and cool point distribution map	Week 2
The first 10 chapters	Each chapter is 2500-3000 words, complete the first draft + polishing	Week 3
First draft of the book	200,000-1 million words, complete narrative chain	2-3 months
Polish and finalize	Full book consistency check + polish	3-4 months
Multi-platform release	Qidian/Tomato/KDP/Webnovel	Continuous
Data review report	Weekly reading + comment analysis	Weekly

Quantitative indicators

• Daily update capability: from purely manual writing of 3,000-5,000 words to AI-assisted writing of 8,000-15,000 words
• Complete this cycle: 300,000-word novel shortened from 6 months to 2-3 months, 1 million-word novel shortened from 18 months to 6-8 months
• Quality Index: The average chapter reading completion rate after AI assistance is no less than 90% of the purely manual version
• Income Expectation: The monthly manuscript fee for Qidian contracted authors is ¥2,000-20,000 (depending on the number of subscriptions), KDP e-books are ¥5-50/book, and IP incubation income is more flexible

Acceptance criteria

• [ ] The outline of the book is logically complete and can cover the entire length of the book
• [ ] The text has 10 consecutive chapters with no setting conflicts or OOC characters.
• [ ] Passed the consistency check of writing style (randomly selected 3 chapters with no major style deviation)
• [ ] Successfully published on at least one platform and passed review
• [ ] The reading rate is stable above the average of similar works
• [ ] Complete at least 1 IP derivative (cover/audio book/comic storyboard)

Frequently Asked Questions and Troubleshooting

Q1: I don’t know how to write prompts at all, how do I start?

Start with the simplest "copy and paste": describe the scene you want to write to the AI in vernacular, and add "Please write a 2,000-word chapter in a web style." Don't pursue the prompt technique in the early stage, run through the "write it out" process first. After you become proficient, you can learn structured prompt and system message configuration. It is recommended to start with the free versions of Kimi or Doubao to test the waters and get started at zero cost.

Q2: What should I do if a chapter written by AI is judged as "AI-generated content" by Qidian/Tomato and is not recommended?

This is a hot issue for various platforms in 2025-2026. Coping strategies:

• Directly generate full text without AI, using the mode of "AI writes 30% skeleton + artificially fills 70% flesh and blood"
• After AI is generated, it must be processed to "remove the AI flavor" (manually rewrite the beginning and end + add personal style paragraphs)
• The manual revision rate of each chapter is no less than 60%
• Do not use a large number of APIs with the same IP under the same account.
• Maintain personal characteristics of plot design and character creation, AI is only responsible for grammar and narrative levels

Q3: After writing dozens of chapters, I find that I forgot to recycle the foreshadowing I laid before. How can I remedy this?

Don't panic. Use AI to scan the entire book to find all unclosed foreshadowing lists, and then:

• If the foreshadowing is still "alive" (can be recycled naturally in subsequent chapters): arrange it in the protagonist's next adventure
• If the foreshadowing is completely unrecoverable: arrange for a character to "mention it casually" or "unintentionally reveal it in a certain scene" to end it lightly.
• AI can design 3 "foreshadowing recovery plans" based on the existing plot and choose the most natural one

Q4: I want to publish overseas (English/Traditional Chinese publishing) but I don’t know how to translate. Can AI directly replace me?

AI translation of novels is still in progress, but it will already reach "readable" quality by 2026:

• Publishing online articles overseas: Use DeepSeek or ChatGPT to translate one chapter first, then manually review and then translate in batches - it takes about 30 minutes to manually review each chapter
• KDP Publishing: After translation, Grammarly is used for grammar checking, and Claude is used for native English polishing.
• It is recommended to outsource the translation + polishing process to professional translators, and AI will be responsible for the initial translation and cost reduction (from ¥200/thousand words to ¥30/thousand words)
• Note: The exclusive copyright agreement of the contracted platform may restrict the release of AI translation, and the copyright ownership needs to be confirmed first.

Q5: What should I do if the AI always deviates from my settings when rewriting?

This is the most common technical problem. Solution:

1. Before each continuation of the dialogue, clearly tell the AI "the outline of the current chapter" (don't just give the results of the previous chapter)
2. Fix a "creative constraint" in the message: the protagonist's current level, location, current goal, and undeviable design
3. Take advantage of Claude’s longer context and add the settings document and the last 5 chapters to the conversation
4. Create an "AI-generated forbidden list": which plots/settings cannot be touched by AI (such as "Don't let the protagonist collect a harem", "Don't let the villain whitewash")
5. If the AI deviates continuously, reset the session and start again from the latest details

Q6: I want to write a multi-character ensemble drama, but what should I do if the AI always only focuses on the protagonist?

Explicitly request perspective rotation in Prompt:

This chapter is written using group portraits, please distribute according to the following proportions:
- Protagonist A’s perspective: 40%
- Heroine B’s perspective: 30%
- Villain C’s perspective: 20%
- Supporting character D/E perspective: 10%
Each perspective switching point is separated by ***. Make sure each POV segment has information that advances that character's storyline.

Q7: Who owns the copyright of AI-assisted writing?

This is a legal gray area. Current industry practice:

• You can regard it as "a creation completed by you using AI tools", and the copyright belongs to you
• However, some platform terms have provisions such as "AI-generated content is not eligible for contract signing" or "AI content is not eligible for recommended resources"
• Suggestion: maintain the attitude of "AI-assisted creation, manual-led modification", and do not emphasize the use of AI when submitting articles
• Overseas KDP requires a statement of "AI-assisted creation", but as long as you make enough manual modifications, it is within the acceptable range

Q8: I am stuck halfway through writing my novel. Can AI help me overcome my writing bottleneck?

Can. Feed the current stuck chapter outline + the content of the previous 3 chapters + the last sentence you wrote to the AI. The requirements are:

I'm currently stuck on [specific questions, such as "I don't know how to make the protagonist transition from point A to point B naturally"/"This conversation ended too blandly"].
Please provide 3 completely different options, 300-word examples for each. Don’t give general suggestions, but specific plots that can be implemented directly. 

Advancement and Expansion

Multiplayer collaborative studio mode

When individual writing capabilities are expanded to a studio team (2-5 people), the workflow upgrades:

• Division of labor model: One person is responsible for the outline (AI assistance) + two people are responsible for the first draft generation (DeepSeek batch) + one person is responsible for polishing and verification (Claude review) + one person is responsible for data operation
• Knowledge Base Sharing: Use Feishu Documents/Notion to create a shared setting library, and manage the version of all AI-generated content in a unified manner
• Quality Access Control: Each chapter must pass three levels before output (AI logic detection → AI writing style detection → manual final review)
• Output Expectation: The daily output of the studio can reach 30,000-50,000 words, and the monthly output is 500,000-800,000 words.

Comic adaptation and cross-form IP incubation

• Character drawing: Use Midjourney to generate character concept drawings, and then use AI to continuously generate storyboard sketches that keep the face consistent.
• Comic storyboard: Input novel chapters into the AI comic tool to generate storyboard scripts, manually adjust them and hand them over to the artist.
• Short play script: Use AI to convert novel chapters into short play script format (3-5 minutes per episode, including scene/dialogue/shot prompts)
• Dynamic Comics: AI generated dubbing script + ElevenLabsMultiple character voice generation + AI animation generation

AI audiobook production

• Character Voice Cloning: Create your own voice for each main character with ElevenLabs’ Voice Design
• Audiobook Chapter Dividing: AI automatically marks chapters into narrative paragraphs and dialogue paragraphs, and assigns different timbres
• Dubbing script optimization: AI converts written language into spoken dubbing script (remove written connectives, adjust word order)
• Release channel: Himalaya/Lazy Man Audiobook/WeChat Audiobook/Audible
• ROI reference: It takes about 3-5 days to produce an AI audiobook for a 500,000-word novel, the cost is ¥500-2000, and the monthly passive income after going online is ¥500-3000.

Overseas translation and global distribution

• English version: AI initial translation (DeepSeek/ChatGPT) + Grammar correction (Grammarly) + English native polish (Claude)
• Traditional Chinese version: Simplified and Traditional conversion + AI terminology unification (terminal differences between the two sides of the Taiwan Strait such as "software/software" and "video/video")
• Southeast Asian languages: Thai/Indonesian/Vietnamese initial translation (ChatGPT/GPT translation) + localization review
• Webnovel/GoodNovel: Upload the English AI translation draft directly and publish it after manual proofreading
• Amazon KDP: Directly exported in EPUB format, the cover is generated with Midjourney, and the description is written with AI marketing copy

AI powered writing toolbox extension

• AI automatic naming: batch generation of name pools for character names/place names/moves
• AI Combat Power Calculator: Helps maintain the upgrade balance in the power system
• AI Dialogue Polisher: A polishing model specially designed for web text dialogue scenes
• AI Rhythm Analyzer: Analyze the reader’s attention curve at the beginning/middle/end of each chapter
• AI Outline Visualization: Convert text outlines into mind maps and Gantt charts