LitRPG & Progression Fantasy: AI Imitation Test on Royal Road

In this article 9 sections

If you want to try writing progression fantasy or LitRPG in English, you can pick a popular book from the public library right in the AI Imitation section on the Maliang Writer English homepage as a style reference, letting the agent read reference chapters before drafting an original opening and continuation chapters. In our hands-on test, both models built a crafter protagonist, skill advancements, and ledgers or progress tracking within three chapters—though finalizing new rules, verifying character knowledge order, and checking continuation chapters for literal overlap still require the author's own oversight.

Key Takeaways

• Fantasy and progression tags appear widely across the top 50 library titles: Among the top 50 Royal Road works by views in the Maliang Writer public library, alongside high-frequency tags like Adventure, Fantasy, and Action, titles also carry tags such as LitRPG (26 books), Portal Fantasy / Isekai (21), Progression (17), Slice of Life (13), and Cultivation (12); a single serial like The Runesmith carries 5 tags at once.
• AI Imitation reads before writing and includes a literal overlap check: When you pick a reference title on the Maliang Writer English homepage (/en/) and send Imitate prose, the agent reads several chapters of the reference book before drafting and runs a literal overlap check against it afterward; in our test, Carbon confirmed zero literal phrase overlaps against the reference chapters in its continuation reply, whereas Sonnet ran the overlap check only on Chapter 1.
• Three chapters are enough to set up a crafter lead and progression framework: Using The Runesmith as a style reference, Gemini 4 Carbon wrote the 3-chapter, 14,303-word The Fault-Tinker, while Claude Sonnet 5.5 wrote the 3-chapter, 10,551-word The Thermowright; the former established currency exchange rates and a mana-and-coin ledger, while the latter tracked a fault-resolution counter and skill progress in its system panel.
• Both models logged progression ledgers or newly coined rules: In its continuation reply, Carbon listed the class, five skill levels, and currency exchange rate, and created a new Ledger-and-Progression.md file to track mana and money; Sonnet noted in its reply that the newly added rules were provisionally set by the AI and open for the author to revise. If you plan to write a long-running serial, our editorial suggestion is to try compiling this kind of data into your reference notes for easy lookup.
• Timeline fixes and ongoing overlap checks still rely on the author: During our test, Sonnet caught and fixed a continuity error during its own review pass—where the protagonist called the young kitchen girl by her name, Wren, in Chapter 2 before ever meeting her—and did not re-run the literal overlap check on Chapters 2 and 3, so a full human read-through is still needed before publishing.

How Common Are Progression Fantasy and LitRPG Among Popular Titles?

First, a note on the data source: the breakdown below comes from the top 50 Royal Road works by views in the Maliang Writer public library (sampled on 2026-11-12, using tags entered by the authors on Royal Road). It reflects the catalog sample in the Maliang Writer public library, not an official Royal Road chart.

Across those 50 works, individual tags appear in the following book counts:

• Genre and story tone: Adventure (44 books), Fantasy (40), Action (34), Comedy (17), Slice of Life (13), Magic (11).
• Progression and portal mechanics: LitRPG (26 books), Portal Fantasy / Isekai (21), Progression (17), Cultivation (12), Reincarnation (6).
• Lead archetypes and subgenre setups: Male Lead (18 books), Anti-Hero Lead (9), Female Lead (9), Non-Human Lead (4), Kingdom Building (3), Dungeon Core (3), Time Loop (3).

Titles near the top of our public library's popular list include Chaotic Craftsman Worships The Cube, The Runesmith, Everybody Loves Large Chests, Super Supportive, Savage Divinity, Path to Transcendence, Stray Cat Strut, Beware Of Chicken, Mother of Learning, Dungeon Diver, and Delve.

As the numbers show, Adventure (44 books), Fantasy (40), and Action (34) lead in frequency, while the list also spans works tagged LitRPG (26), Portal Fantasy / Isekai (21), Progression (17), Slice of Life (13), and Cultivation (12). A single book routinely stacks multiple tags—for example, The Runesmith, which we picked as our style reference for this test, carries five tags at once: LitRPG, Portal Fantasy / Isekai, Kingdom Building, Male Lead, and Slice of Life.

Four Elements Across the Library Tags and Test Samples

Looking strictly at the tag breakdown of those top 50 library titles and the sample prose generated by the two models, four concrete elements stand out in this test:

1. Visible growth (levels, skills, and class advancements)
   Out of the top 50 library books, 26 are tagged LitRPG and 17 are tagged Progression, and our generated samples likewise break power growth into clear milestones. In The Fault-Tinker by Gemini 4 Carbon, the protagonist levels up skills—Damping Calibration from Level 1 to Level 2 and Tinkering from Level 2 to Level 3—and advances from Tier 0 Scrapper Apprentice to Tier 1 Fault-Tinker; in The Thermowright by Claude Sonnet 5.5, the protagonist unlocks Thermal Sense in Chapter 2 and pushes Leak Finder progress to 2/10 in Chapter 3.

2. Status prompts and system panels
   The two test samples use bracketed status lines and system panels, respectively, to display character state. The Fault-Tinker holds off on bracketed status lines throughout Chapter 1, waiting until Chapters 2 and 3 to pop single-line alerts such as [Damping Calibration: Level 1 -> Level 2] and [Class Updated: Tier 1 Fault-Tinker]. The Thermowright uses gray system text that fades without a trace, with a panel logging the protagonist's physical and past-life mental age as Age 10 (38), a +2 Perception stat bonus, and Stubborn Bastard, a skill triggered for 1 hour whenever someone tells him "no."

3. Resource and money tracking
   Both test samples log coin, mana, or progress numbers in either the chat reply or the story panel: The Fault-Tinker sets out a currency exchange rate of 1 silver crown = 100 copper bits in its continuation reply and creates a dedicated ledger file, Ledger-and-Progression.md, to track mana values and money; The Thermowright titles Chapter 2 Cheapest First as an older kitchen worker drags a big pot onto the range, shows the protagonist hiding a list scribbled on flour sacking alongside a rag-wrapped cup inside a gap in the woodstack in Chapter 3, and logs Fault resolved: 2 and 2/10 skill progress on his panel.

4. Hands-on crafter and repair protagonists
   Our reference serial The Runesmith carries tags like LitRPG, Kingdom Building, and Slice of Life, and both generated samples similarly center on hands-on trade and repair leads: Callum Vance in The Fault-Tinker sits inside a shop patched with corrugated tin on the roof, keeping watch over an oak workbench piled with corroded brass fittings and stripped iron bolts, plus a cedar quenching bucket; Mac Dunleavy in The Thermowright is a modern HVAC tech who goes from servicing an aging 1998 condensing unit on a 19-degree roof in his past life to working as a ten-year-old kitchen scullion after reincarnation.

Hands-On Test: Imitating The Runesmith on the English Homepage

On 2026-11-12, we went to the AI Imitation section on the Maliang Writer English homepage (/en/), chose Royal Road as the source, clicked the ongoing serial The Runesmith from the popular list (under copyright, so no original text from the reference book is quoted here), and clicked Imitate prose to send. Here is the exact English prompt generated automatically by the homepage:

Use the writing style of [Royal Road:…The Runesmith…] as a reference and write a prose passage for me in the same style. Don't copy it—make it original, with your own twists and changes.

Once Chapter 1 finished generating, we sent one follow-up message in English in the same chat session (verbatim):

Continue with chapters 2 and 3. Keep the voice, the progression system and every number established in chapter 1.

We ran one three-chapter project each using Gemini 4 Carbon (thinking: medium) and Claude Sonnet 5.5. During testing, we verified every chat turn and prose-generation log item by item to confirm that the tested model itself handled both conversation and drafting; the English prose received zero human edits, and length was tallied by English word count. Before drafting, the agent reads several chapters of the reference work, and afterward it runs a literal overlap check against the reference chapters.

Sample 1: Gemini 4 Carbon — The Fault-Tinker

• Book title, chapter titles, and word counts: The AI titled the book The Fault-Tinker. Its three chapters are Chapter 1 - Bad Angles and Cheap Brass (4,317 words), Chapter 2 - Lead Lining and Shift Change (4,816 words), and Chapter 3 - Silver Paste and Buried Iron (5,170 words), totaling 14,303 words.

• Time and credits: Chapter 1 took 1,586 seconds and used 827.0 credits; continuing with Chapters 2–3 took 2,074 seconds and used 777.72 credits.

• Opening excerpts: Chapter 1 opens with cold soot-rain, corrugated tin roof patches, and scrap hardware on the workbench:

  Cold soot-rain pattered against the corrugated tin patching the roof of the shop. Inside, a single tallow lamp guttered on the corner of the scarred oak workbench, throwing greasy yellow light across a pile of corroded brass fittings and stripped iron bolts.

  Chapter 2 opens by counting the seconds for ripples in a cedar quenching bucket to settle before cutting to a clicking silver spike and a deep tremor in the floor:

  The concentric ripples in the cedar quenching bucket took four seconds to die out, and right on the fifth second, the matte-silver spike on the oak planks clicked again. Tick. Three heartbeats later, the dirt floor answered with that same slow, heavy thrum from straight down through the bedrock.

  Chapter 3 opens with the protagonist carrying a scorched guild frame in his bag and ducking off the main thoroughfare into a narrow lane:

  Callum stayed off High Street completely, cutting south below the timber yards into the narrow cobbles of Ash-Gutter Lane. Going up the main hill meant walking past Gilded Spire Artificery's front shutters in broad daylight with a scorched guild frame in his bag and Garrett's muddy tracks still fresh down at Pit Four.

• Status prompts: Chapter 1 contains no bracketed status lines; Chapters 2–3 introduce at least four verbatim prompts: [Damping Calibration: Level 1 -> Level 2], [Tinkering: Level 2 -> Level 3], [Class Advancement Available: Tier 0 Scrapper Apprentice -> Select Tier 1 Class], and [Class Updated: Tier 1 Fault-Tinker].

• AI's self-logged progression ledger: In its continuation reply, Carbon listed the protagonist's current class and skills—Class & Skills: Tier 1 Fault-Tinker; Fault Resonance (L2), Damping Calibration (L2), Tinkering (L3), Scrap Appraisal (L3), Harmonic Seating (L1).—noted the currency rate Ledger (1 silver crown = 100 copper bits), and stated that it confirmed zero literal phrase overlaps against the reference chapters. It also created a separate ledger file, Ledger-and-Progression.md, to track mana values and money.

• Editorial notes: Every chapter came in above 4,300 words and grew longer from chapter to chapter; Chapter 1 skipped bracketed status lines entirely before delivering skill level-ups and a Tier 1 class advancement across Chapters 2–3. Working from the English homepage, Carbon also wrote its chat replies in English.

Sample 2: Claude Sonnet 5.5 — The Thermowright

• Book title, chapter titles, and word counts: The AI titled the book The Thermowright. Its three chapters are Chapter 1 - Of All the Ways to Go (3,151 words), Chapter 2 - Cheapest First (4,033 words), and Chapter 3 - Where the Heat Goes (3,367 words), totaling 10,551 words.

• Time and credits: Chapter 1 took 246 seconds and used 238.47 credits; continuing with Chapters 2–3 took 1,246 seconds and used 414.06 credits.

• Opening excerpts: Chapter 1 opens with the protagonist up on a 19-degree roof facing an aging condensing unit from 1998:

  Mac Dunleavy had been on the roof for forty minutes. It was nineteen degrees, and the condensing unit in front of him was older than half the people who worked underneath it. 1998.

  Chapter 2 picks up with fading gray system text contrasted against the soot smudged on his nose:

  The gray text faded without a trace. Not even a smudge, which was more than he could say for the soot on his nose.

  Chapter 3 opens with the protagonist wrapping a cup in a rag and stashing it in a gap in the woodstack beside his list on a scrap of flour sacking:

  He got the cup into a rag, rolled it twice, and shoved the bundle deep into the gap in the woodstack, next to the scrap of flour sacking he'd started his list on. Nobody in this kitchen looked behind the kindling.

• Status prompts: Protagonist Mac Dunleavy reincarnates from a modern HVAC technician into a kitchen scullion (Age 10 (38)), starting with system panel skills Leak Finder and Stubborn Bastard (triggered for 1 hour when told "no"); Chapter 2 unlocks Thermal Sense (+2 Perception); in Chapter 3, the panel logs Fault resolved: 2 and Leak Finder progress at 2/10.

• AI's self-logged progression ledger: In its chat reply, Sonnet noted that the newly added rules were its own provisional inventions that the author could modify; during its review pass, it fixed a continuity error (in Chapter 2, the protagonist had called the young kitchen girl by her name, Wren, before ever meeting her); and the literal overlap check was run only on Chapter 1, not repeated on Chapters 2 and 3.

• Editorial notes: Individual chapters ranged from 3,151 to 4,033 words, pivoting from a modern rooftop HVAC call in Chapter 1 into an otherworldly kitchen after reincarnation. One quirk worth noting: even though we ran the session on the English homepage and Sonnet wrote the novel prose in English, its conversational replies in the chat box came back in Chinese.

Keeping Numbers Consistent: Letting the AI Keep a Ledger

Across this three-chapter run, the two models tracked stats and system rules in different ways: Gemini 4 Carbon both summarized the current Tier 1 class, five skill levels, and silver-crown-to-copper-bit exchange rate in its continuation reply and spun up a dedicated Ledger-and-Progression.md file for mana values and coin; Claude Sonnet 5.5 incremented its fault counter and skill progress inside the story's system panel while noting in chat that new rules were provisional and open to edits.

Our test only ran for three chapters, and Ledger-and-Progression.md was created on Carbon's own initiative rather than requested in our prompt. If you plan to keep a serial going over many chapters, our personal suggestion as editors is to keep instructing the AI during continuations to preserve previously established numbers—and to compile (or ask the AI to compile) character levels, skills, exchange rates, and resource changes into reference notes that you double-check by hand after every chapter. For a closer look at how different models hold onto stat blocks across multi-chapter runs, check out our Hands-On Test of AI Stat Consistency in LitRPG.

Where AI Helps and What the Author Needs to Check

Based on what happened across both test samples, here is what the AI handled in this run alongside five practical editorial suggestions on where authors should step in:

1. In our test, the AI grounded openings in concrete tools and physical spaces; we suggest authors steer the longer-term plot arc: Both samples opened with tactile workshop and trade details—corroded fittings on an oak bench and timed ripples in a cedar quenching bucket in one, an aging rooftop condensing unit and supplies stashed behind kitchen kindling in the other. If you continue past the opening chapters, try mapping out where the protagonist's broader plot is headed yourself ahead of time.
2. In our test, the AI invented starting skills and exchange rates on its own; we suggest authors lock down system rules early: Carbon introduced class tiers and coin conversion rates, while Sonnet invented a conditional trigger skill and explicitly noted in chat that new rules were provisional for the author to adjust. After reading your first draft, try confirming in chat right away which mechanics you want to keep or rewrite.
3. In our test, single chapters ran from over 3,000 to over 5,000 words; we suggest reading through to verify character knowledge order: Chapter lengths ranged between 3,151 and 5,170 words, and Sonnet caught and fixed an error during review where the protagonist used the kitchen girl Wren's name in Chapter 2 before learning it. Given how long these chapters run, try giving every draft a full human read-through before finalizing, paying special attention to who knows what and when.
4. In our test, Chapter 1 ran a literal overlap check against the reference book; we suggest keeping an eye on overlap checks in later chapters: Carbon confirmed zero literal phrase overlaps against the reference chapters in its continuation reply, whereas Sonnet ran the literal overlap check on Chapter 1 but did not run it again on Chapters 2 and 3. During continuations, try checking whether each new chapter ran its overlap check and reviewing originality yourself.
5. In our test, models summarized progress or created a ledger file; we suggest cross-checking those numbers against the prose: Carbon listed skill levels in chat and created Ledger-and-Progression.md, while Sonnet tracked fault counts and skill progress inside the panel. Before starting each new chapter, try cross-checking those summaries or ledger files directly against what actually happened in the chapter text.

Want to Try It Yourself?

If you want to try generating an English progression fantasy or LitRPG opening and continuation chapters with AI Imitation, follow these steps:

1. Open the Maliang Writer English homepage (/en/) and go to the AI Imitation section.
2. Select Royal Road as the source, then pick a reference book from the popular list whose tags match what you have in mind (such as titles tagged LitRPG, Progression, or Slice of Life).
3. Click Imitate prose to send the default English prompt, then wait for the agent to read several chapters of the reference work, draft an original Chapter 1, and run a literal overlap check against the reference book.
4. Read through Chapter 1 to check the protagonist's background, the English prose voice, and the starting mechanics; if any rules need tweaking, spell out your changes right in the chat.
5. In the same chat session, send a continuation prompt: Continue with chapters 2 and 3. Keep the voice, the progression system and every number established in chapter 1. Once generated, we suggest manually cross-checking any new levels, skills, and resource numbers against the earlier chapters.

FAQ

Q1: What is progression fantasy?

Among the top 50 titles in our public library, 17 carry the Progression tag. In our editorial view, it focuses on stories built around characters steadily growing stronger or more capable over time; in our two test samples, that takes the concrete shape of skills leveling from Level 1 to Level 2, a Tier 0 apprentice advancing to a Tier 1 class, and skill progress ticking up to 2/10 on a panel.

Q2: What is the difference between LitRPG and progression fantasy?

In our tag breakdown of the top 50 library titles, LitRPG (26 books) and Progression (17 books) appear as separate tags. In our editorial view, LitRPG typically surfaces quantized stats explicitly—like the bracketed status prompts, stat bonuses, and skill panels in our test samples—whereas Progression refers more broadly to stories structured around advancing capability and power.

Q3: How much status-panel text is the right amount?

There is no single rule, and our two test samples illustrate two different approaches: Carbon's sample omitted bracketed status lines completely in Chapter 1 before dropping at least four single-line skill and class alerts across Chapters 2 and 3, while Sonnet's sample updated panel counters after unlocking skills and resolving faults. Our suggestion is to tell the AI in chat how light or frequent you want prompts to feel for your pacing.

Q4: Does AI style-matching plagiarize the reference book?

The default AI Imitation prompt explicitly instructs the model not to copy the reference book and to make the prose original with its own twists, and the agent runs a literal overlap check against the reference chapters after writing. Even so, in our test Sonnet ran that check only on Chapter 1 and did not repeat it for Chapters 2 and 3; our personal advice is not to rely solely on automated checks, and to read through the draft yourself before publishing to verify both plot and phrasing originality.

Q5: How long does one chapter take, and how many credits does it cost?

Generation time and credit usage vary by model and chapter length. In this test, Carbon took 1,586 seconds and 827.0 credits for Chapter 1, plus 2,074 seconds and 777.72 credits for Chapters 2–3; Sonnet took 246 seconds and 238.47 credits for Chapter 1, plus 1,246 seconds and 414.06 credits for Chapters 2–3, with individual chapters landing between roughly 3,100 and 5,100+ words.

Q6: Can AI-written stories be posted on Royal Road?

For disclosure rules on posting AI-assisted or AI-generated fiction on Royal Road, see our AI Policy Guide for Overseas Web Novel Platforms. Before posting, we recommend reading through the entire manuscript yourself to verify levels and resource ledgers, check character knowledge timelines, confirm overlap checks, and take full ownership of the final draft.

Conclusion

When you pick a Royal Road library title as a style reference in the AI Imitation section on the Maliang Writer English homepage, the agent reads reference chapters first before drafting an original English opening and continuation chapters—supplying skill advancements, panel stats, or ledger files along the way. As our three-chapter test also showed, generation can come with AI-invented mechanics, character-timeline slips, or skipped literal overlap checks on follow-up chapters; if you try it for your own serial, pairing reference ledgers with a chapter-by-chapter human read-through will help you keep tight control over the finished manuscript.

Related Reading:

• Hands-On Test of AI Stat Consistency in LitRPG
• AI Policy Guide for Overseas Web Novel Platforms
• Gemini 4 Carbon Fiction Writing Hands-On Test