Maliang Writer
Writing Craft15 min read

Characters Collapsing at 500K? Manage 100 Characters with AI

Past 500,000 characters, keep 100 characters consistent: state cards, relationship tracking, contradiction checks, and an ensemble case study.

The Maliang Writer Team · Maliang Writer

In this article 6 sections

Why Your Characters Are Bound to Fall Apart

It’s not that you aren’t trying; the human brain really can’t remember this much.

A 500,000-character web novel usually has 60–100 named characters. Each character has their own:

  • Basic attributes (personality, appearance, catchphrase, cultivation level)
  • Dynamic state (current location, items held, recent events experienced)
  • Relationship network (relationships with other characters, the events that triggered changes)
  • Growth arc (what kind of person they start as and become, the key turning points)

Remember all of it in your head? Problems start by 300,000 characters. By 500,000 they are certain.

The most common types of “continuity errors”:

  1. State regression: the protagonist broke through to the Golden Core stage in the last volume, and this volume he is suddenly back at the Foundation Establishment stage
  2. Relationship contradictions: A and B fell out in chapter 15, and in chapter 30 they chat like good friends again
  3. Personality drift: a taciturn character suddenly becomes talkative with no setup
  4. Lost items: an ancient divine artifact is obtained in chapter 20 and never mentioned again

Readers spot these problems far faster than you think. Web-novel readers follow updates seriously, and will flip back through earlier chapters to “call out fakes”.

Why Traditional Solutions Don’t Work Well

Most authors manage characters with three kinds of tools:

1. Character cards (Word/Notion documents)
→ Problem: later on you’re too lazy to update them, and the information is always out of date

2. Mind maps
→ Problem: they can only show static relationships and can’t track dynamic changes

3. Excel spreadsheets
→ Problem: inconvenient to search, and you have to switch by hand between them and your writing tool

The core problem of these approaches is: the maintenance cost is too high, and they are separate from the writing process.

You can’t update the state table of 100 characters after every chapter. Writing until three in the morning to make the update is tiring enough; who has the energy to fill in Excel?

The Core Idea of AI Character Management

When managing characters with AI, the point isn’t to have AI “design characters” for you, but to have AI “remember characters” for you.

Specifically, AI does three things in character management:

Thing 1: Automatically Extract and Update Character States

After each chapter you write, AI automatically extracts character state changes from the body text:

  • A character appeared → update “most recent appearance chapter”
  • A character broke through in cultivation → update “current cultivation level”
  • A character obtained a weapon → add it to the “items held” list
  • A character came into conflict with another → update the relationship status

In Maliang Writer, this feature is implemented through the knowledge graph engine. You write the body text, and the graph updates automatically.

Thing 2: Conflict Detection

When AI generates or checks new chapter content, it automatically compares the current text against the existing character states:

  • “You wrote Zhang San holding the Chixiao Sword, but according to the graph records, the Chixiao Sword was given to Li Si in chapter 45”
  • “You wrote the protagonist traveling from City A to City B in just one day, but according to the geography settings the two cities are 3,000 li apart”
  • “You wrote Wang Wu saying ‘Hello, senior brother’, but he has no martial-sibling relationship with that person”

This kind of detection doesn’t require AI to understand the plot, only to do fact-checking. Like a conscientious editor checking the story bible word by word.

Thing 3: Visualizing the Relationship Graph

The relationship network among 100 characters is too complex to describe in words. AI can help you generate a visual relationship graph:

  • Nodes = characters
  • Lines = relationships (master-disciple, hostile, lovers, allies)
  • Colors = camps or factions
  • Node size = frequency of appearance

One diagram shows the whole story’s cast layout. This is especially useful for multi-thread narratives: you can immediately see which characters haven’t appeared for too long and which relationship threads have been neglected.

Character relationship management

In Practice: How to Build a Character Management System

Step 1: Build a Character File Template

Every character needs at least the following information:

Basic information (filled in at creation, rarely changes)

  • Full name and aliases
  • Age / appearance features
  • Personality keywords (3–5, such as “steady / loyal / stubborn”)
  • Signature catchphrase or speech habits
  • Core motivation (why they do what they do)

Dynamic information (updated with the plot)

  • Current cultivation / power level
  • Current location
  • Important items held
  • Major events experienced recently
  • Relationship status with other characters

Step 2: Set Update Rules

Not every change needs recording. We suggest tracking only state changes that “affect the plot”:

  • Cultivation breakthrough → record
  • Changed into ordinary clothes → don’t record
  • Allied with a character → record
  • Passed through an unrelated small town → don’t record

In Maliang Writer, the consistency agent automatically decides which changes need to be entered, with no manual filtering on your part.

Step 3: Run a Regular “Global Checkup”

After each volume (or every 100,000 characters), do a global character checkup:

  1. Which characters haven’t appeared for more than 5 chapters? Was that deliberate or forgotten?
  2. Are there characters whose personality descriptions contradict each other?
  3. Are there “broken” lines in the relationship graph (two characters in conflict that was never resolved)?
  4. Are the growth arcs coherent? Is there any “leapfrog growth”?

List these questions as a checklist and check them one by one. AI can generate this checkup report for you automatically.

Pitfalls That Come Up Often

Pitfall 1: Character Settings That Are Too Detailed

The mistake beginners make most easily: writing out a character’s past lives and three generations of family in detail, then finding they’re never used when writing the body text.

Character settings should be only as detailed as needed. Core motivation, personality through-line and key relationships are all you need. The rest can be filled in when you reach it.

Pitfall 2: Neglecting Supporting Characters

Many authors only seriously manage the protagonist and a few core supporting characters. But what most easily causes “continuity errors” are exactly the peripheral characters who appear rarely but pop up occasionally.

Even a passerby who has appeared three times will be remembered by readers. If the first time he was a steamed-bun seller and the third time he is a blacksmith, someone will point it out in the comments.

Pitfall 3: Not Recording the Timeline

This is the most hidden but most fatal pitfall. Without recording the chronological order of events, the consequences are:

  • Events enough for three days happen within one day
  • A character appears in two places at the same time
  • Season and weather descriptions contradict each other

We suggest maintaining a simple timeline separately from character management. It needn’t be accurate to the hour, but you should at least know whether “this happened before or after that”.

Summary

Character management is not an extra burden on creation, but the infrastructure of long-form writing. Just as constructing a building needs a foundation, a long novel of over 500,000 characters must have a reliable character management system.

AI’s greatest value isn’t helping you design better characters, but helping you faithfully remember every one of them. Only if you remember them won’t you mix them up. Only if you don’t mix them up will readers trust your story.

Back to all articles