The organizing insight
It isn't one market. It's an archipelago.
The famous <10% same-age-female overlap (贾毅 / Crazy Maple, 2023) — which the fact-check narrowed to Chapters' own four IF titles — means these apps run the same engine but their content library + ad-creative + art register each select a different psychographic of the "same" woman. Consequence: a new format is audience acquisition, not cannibalization — you don't fight incumbents, you settle a new island.
Chapters / Episode / Choices. The historical center — but the format shrank ~36% in 2022; monetization broke (IDFA), the appetite didn't.
Love & Deepspace: ~$933M lifetime, Year 2 > Year 1. One husbando gacha out-earns the whole surviving IF category.
ReelShort / DramaBox. $2.98B IAP in 2025. The paying core is older than everyone assumes — Hallmark's audience, mobile-native.
Zeta (65% F, profitable), Talkie, Xingye. Character.AI is ~80% male — the "80% female" figure was a misquote the fact-check caught.
Dear U bubble (~41% margin on scripted idol DMs), 谷子 merch (~¥169B). Story apps sell ¥0 into this.
AO3 (~125M pageviews/day) — only ~12% straight, ~27% ace. The "default straight self-insert reader" is statistically the exception.
Who she really is · Round 5
Not lonely. Depleted.
The iteration
Ten rounds, each drilling the last
Every round adversarially fact-checked the previous one — the verify layer caught several fabricated statistics and a misquoted gender split before they reached the founder.
The fandom archipelago
21 tracks map the audience, the money, the <10% overlap, the whitespace.
The wedge + the agency rule
Operator teardowns of the current winners; where the audience went.
Build verdict + 90-day plan
Feasibility, identity-consistency, COGS, the keepsake WTP case.
After a major entered
Naver byUs shipped the near-identical wedge one week earlier.
Who she really is
Ethnographic depth — in her own words, incl. the ones who bounced.
GTM + ethics operating spec
Growth loop, payments, cohort, the grief line, bridge-not-cave.
Revenue + defensibility
The no-gacha model — and the crux: the architecture is pre-rendered, not live.
Forward view + red-team
2027–30 forecast + an adversarial fleet built to kill the thesis.
Survival playbook + the call
Persistence retrofit, guided-freedom hybrid, taste solve; every "better door" fails harder.
Tripwire sweep
6 dated signals that could re-price the verdict.
The integrated call
NARROW-GO — chosen, not defaulted
A defensible lifestyle-scale craft business (~$1.5–6M/yr ceiling), not a venture — surviving only if the moat relocates onto taste + human-authored provenance + a persistence retrofit + the midlife-adult register, and only if it clears a pre-registered kill-gauntlet before real spend. The wedge: self-insert, painterly-adult, voice-first, pre-rendered, no-gacha, "playable romantasy," not "AI companion."
- 01Video stays pre-rendered, never live-per-user — the solvency mechanism.
- 02Web-first checkout from day one (~94¢/$ vs 70¢ in-app).
- 03Charge on the completed arc + keepsake, never sessions — the low-harm, legally-durable metric.
- 04A persistence retrofit (day-counter + scheduled letter + callbacks) beats the leaky bucket.
- 05Stay out of "companion" — authored branches, no free-form chat, adults-only.
What to do next · decision ≤ Dec 2026 · ≤ $25k
A pre-registered 4-gate kill sequence
Nine rounds produced zero primary user contact. The next unit of truth isn't researchable — it's a $1k taste test. Each gate has a pre-specified kill rule; ambiguity resolves by tiebreaker, never a re-run.
The closest competitor · Naver byUs
A validation gift that can't follow you
byUs (Naver Webtoon, launched July 1 2026, Korea-only, in-app) proves self-insert monetizes hard — 41% paid-message share on its predecessor — while structurally fencing itself off from your differentiators. It's canon-locked by legal necessity (six webtoon firms filed criminal complaints against the user-OC app Zeta), so it cannot let users own their own character without becoming a different company.
- ✓Self-insert persona — byUs has it (text, no avatar)
- ○Interactive video — open (byUs is text + buggy static images)
- ○Voice ritual — open
- ○Web-first / Western — open (Korea-only, in-app)
- ○User-owned original characters — open (canon-locked)
- ○Exportable keepsake — open (rewards are in-app only)
Steal: the affinity → reward-image loop is the monetization engine. But byUs's rewards are non-exportable — so it has no organic growth loop. A shareable self-insert moment is exactly the loop it can't build.
Avoid: the pay-before-fun paywall (~4 free messages gates users out before the "wow"), and shipping the headline visual feature buggy (café rendered as an office) — for a video-first product, the visual is the product.
The record
12 cited, fact-checked documents
Every round's Section B ledger corrected the prior round's numbers. Full markdown lives in docs/research/.