信念收割 AI 全景图 · 一次身份的许诺 · 2026-07 · 防御与识别地图Belief Harvesting × AI · One Promise of Identity · Jul 2026 · A Defence & Recognition Map
你以为在买梯子,其实你是梯子
You are the ladder
这个产业卖的是「向上流动的幻觉」:你以为在买一把梯子,其实你就是那把梯子——AI 没有发明新骗局,它把每个旧环节的边际成本砸到接近零
This industry sells the illusion of upward mobility: you think you are buying a ladder — you are the ladder. AI invented no new con; it crushed the marginal cost of every old step to near zero
本图是什么:一张防御与识别地图。它与诈骗图分工——诈骗图管违法区的刑事骗局;本图管「合法外衣区」的光谱:从高价但合法,到接近犯罪,交界处站着游走两界的资金盘。最小交易单元不是一门课,是「一次身份的许诺」。What this map is: a defence and recognition map. It splits duties with the fraud map — that one covers the criminal zone; this one covers the lawful-veneer spectrum, from pricey-but-legal to nearly criminal, with money games straddling the border. The atomic transaction is not a course but one promise of identity.
本图不是什么:不是运营手册。只提供识别框架,不提供任何可操作的收割/运营方法;所有点名个案均以公开的行政处罚、司法判决或权威媒体调查为准。What it is not: an operations manual. It offers recognition frames only, no actionable harvesting methods; every named case rests on public penalties, judgments or authoritative investigations.
主脊:一个参与者被收割的九阶段生命周期(引流→钩子→身份重构→高价转化→圈层绑定→角色反转→深度收割→沉没或觉醒→维权困境),角色反转是全图枢纽——买家变下线的那一步,就是传销结构的法律定义时刻。三条结构带:法律光谱、AI 成本结构改写、治理与出清。判断层四条元规则,最硬一条:查官方收入披露与退款率,胜过听任何见证。姊妹分工:玄学收割链→mystic,认知偏误→mind,成长焦虑→growth,加密资金盘→crypto。
The spine: a participant’s nine-stage harvesting lifecycle (funnel → hook → identity rebuild → high-ticket conversion → circle binding → role reversal → deep harvest → sinking or waking → the redress trap); role reversal is the hub — the step from buyer to downline is the legal definition-moment of a pyramid. Three bands: the legal spectrum, AI’s cost rewrite, governance and culls. Four meta-rules, the hardest first: check official income disclosures and refund rates over any testimonial. Siblings: the mystic harvest chain → mystic, bias → mind, growth anxiety → growth, crypto money games → crypto.
35 亿
「创造丰盛」灵修案非法聚敛金额,涉 1 万余人 28 省,2025 年一审 8 名骨干获刑(判决 A)——「身份的许诺」可以贵到这个量级The «Creating Abundance» retreat scheme took ¥3.5B from 10k+ people across 28 provinces; eight leaders sentenced in 2025 (judgment, A) — a promise of identity can cost this much
≤$1000
FTC 分析 70 家 MLM 的 2023 收入披露:多数参与者年收入不超过 1000 美元,至少 17 家多数参与者分文未赚(A)——行业自己签字的诚实层FTC’s read of 70 MLM income disclosures for 2023: most participants made ≤$1,000 a year, and at 17 firms most made nothing (A) — the honesty layer the industry itself signed
1.75 亿
一位「AI 导师」三年卖课收入的媒体测算(C;另一口径:两门 199 元课售 14 万份约 2786 万)——AI 焦虑是韭菜经济的新石油A media estimate of one «AI mentor’s» three-year course revenue (¥175M, C; another basis: 140k copies of two ¥199 courses ≈ ¥27.9M) — AI anxiety is the leek economy’s new oil
707 万人
一美妆品牌被认定传销时的会员规模,涉传收入近 92 亿、罚没约 2098 万(行政处罚 A)——角色反转的工业化规模Membership of one cosmetics brand when ruled a pyramid: 7.07M people, ~¥9.2B in scheme revenue, ~¥21M confiscated and fined (administrative ruling, A) — role reversal at industrial scale
⚠️ 口径裁判(先读):① 个案金额存在「实际获利」与「累计流水」两种口径的巨大差异——善心汇「非法获利 25 亿」(官方)≠「涉案 1046 亿」(媒体);PlusToken「148 亿」(法院按历史低点)≠「400 余亿」(媒体按查获总量);本页就地标注,不可混用;② 广为流传的「MLM 99% 参与者亏损」不是 FTC 官方结论,而是独立研究者提交 FTC 存档的报告结论,常被误引;③ 「AI 导师」收入 2786 万与 1.75 亿为两个口径的媒体测算,均非司法认定;④ 直销牌照「88/89 家」两口径未交叉验证;思八达 45 万课价为单源;⑤ 洗脑「解冻—改变—再冻结」三阶段为媒体科普概括,非学术一手文献;⑥ 本页与诈骗图分工:此处只管「合法外衣区」光谱,刑事骗局归诈骗侧;⑦ 四份深度研究交叉整理(两份带可核 URL 为主力,约 60 条带源发现);渗透%为编辑估值。
⚠️ Basis rulings (read first): ① case amounts split between «actual gains» and «cumulative flow» — one charity-front scheme’s «¥2.5B illegal gains» (official) ≠ «¥104.6B involved» (media); a crypto pyramid’s «¥14.8B» (court, at historic lows) ≠ «¥40B+» (media, at seizure value); flagged in place, never merged; ② the famous «99% of MLM participants lose money» is not an official FTC finding but an independent researcher’s archived submission, routinely misattributed; ③ the «AI mentor’s» ¥27.9M and ¥175M are two media estimates, neither judicially found; ④ the «88/89 licensed direct sellers» bases are uncrossed; one ¥450k course price is single-sourced; ⑤ the «unfreeze–change–refreeze» model is media shorthand, not primary scholarship; ⑥ this page covers the lawful-veneer spectrum only — criminal cons belong to the fraud map; ⑦ cross-compiled from four deep-research reports (two URL-verified as backbone, ~60 sourced findings); penetration %s are editorial.