夜间经济 · 一小时被安放的夜晚 × AI · 2026-07The Night Economy · An Hour, Safely Spent × AI · Jul 2026

杀死 KTV 和网吧的不是 AI,是智能手机——AI 只是补刀者。它造不出「想聚」的念头,却决定谁能被看见、谁的深夜无人值守、谁向监管报告;而夜晚最贵的东西,正在变成「没有算法」 What killed the KTV and the internet café was not AI but the smartphone — AI only finishes the wounded. It cannot conjure the urge to gather, yet it decides who gets seen, whose small hours run unstaffed, who reports to the regulator — while the dearest thing in the night is becoming «no algorithm at all»

这个行业卖的不是歌、球、水、汤,是「一小时被安放的夜晚」——场地+氛围+服务打包的时间消磨。账本的对照:2025 年持证娱乐场所营收 430.4 亿,只有线上文化娱乐(11033.8 亿)的 1/25(文旅部,A)——价值被屏幕整整抽走了一个数量级。What this trade sells is not songs, cues, water or steam but an hour of the night, safely spent — venue, atmosphere and service bundled into a way to pass time. The ledger’s contrast: licensed venues earned ¥43.04B in 2025 — 1/25th of online culture and entertainment’s ¥1.10T (MCT, A) — a full order of magnitude drained into the screen.
全图趋势四联句:前台更真人,后台更 AI;标准化更无人,情绪化更有人(另两联:流量更算法、体验更反算法)。哑铃正在成形:一端全自助无人化,一端反算法人味化——站在中间的业态最先死The map’s four-line law: more human out front, more AI in back; the standardised goes unmanned, the emotional stays manned (plus: traffic ever more algorithmic, experience ever more anti-algorithmic). A dumbbell is forming — all-self-service at one end, anti-algorithm warmth at the other — and whatever stands in the middle dies first.

主脊是一夜的时间条:黄昏到黎明八个时刻节点,每个节点=「几点钟+生命周期第几幕」(起意→选址→到店→核心消费→加值→买单→复购),灯光语义=渗透度(冷光=AI 深,暖光=人味浓)。附十大业态兴衰灯牌墓园、流量税三方博弈、无人化时段、产品指南(2C/2B)。姊妹分工:夜宵→dine,售票演出→show,棋牌赌边→bet,酒精成瘾→addiction The spine is one night’s time-bar: eight moments from dusk to dawn, each pairing «the hour + the lifecycle act» (urge → choosing → arrival → core spend → upsell → the bill → return), with light as semantics — cold light = deep AI, warm light = human warmth. Plus the neon graveyard of ten formats, the three-way traffic-tax war, the unmanned hours and a product guide (2C/2B). Siblings: late-night food → dine, ticketed shows → show, the gambling edge → bet, alcohol & addiction → addiction.

传统节点Traditional
冷光 · AI 渗透深Cold light · deep AI
暖光 · 人味浓Warm light · human
流量税 · 平台Traffic tax · platforms
监管科技RegTech
1/25
2025 年持证娱乐场所营收(430.4 亿)与经营性互联网文化单位(11033.8 亿)之比(文旅部公报,A)——「价值被线上抽走」一个数字讲完。⚠️430 亿是持证 KTV/歌舞窄口径,切勿当「夜经济总规模」引用Licensed venues’ 2025 revenue (¥43.04B) against online culture’s ¥1.10T (MCT bulletin, A) — value drained online, in one number. ⚠️A narrow licensed-venue basis: never cite it as «the night economy’s size»
−7万 / +11万
一夜之间的死与生:KTV 九年关掉约 7 万家(约 12 万 → 不足 5 万,B);台球相关企业 2024 年 1–10 月新增 113,994 家(艾媒,B)——业态不灭,只是转世Death and birth in one night: KTVs shed ~70k stores in nine years (~120k → under 50k, B); billiards-related firms added 113,994 in Jan–Oct 2024 (iiMedia, B) — formats don’t die, they reincarnate
80–85 vs 50–60%
美团 vs 抖音的团购核销率(⚠️券商研报单源 B−);抖音 80% GTV 来自连锁——货架逻辑利中小店、内容逻辑利连锁:流量税内层战争的分野Meituan vs Douyin voucher redemption (⚠️single brokerage source, B−); 80% of Douyin GTV from chains — shelf logic favours the small shop, content logic the chain: the inner front of the traffic-tax war
48.28%
2025 年消费者弃 KTV 首因=「新兴社交娱乐方式增多」(艾媒,B)——不是 AI,是口袋里的手机与层出不穷的新玩法;客流较 2020 前下滑约七至八成(B)The top reason for quitting KTV in 2025: «too many new ways to socialise» (iiMedia, B) — not AI, but the phone in the pocket and the endless new formats; footfall down ~70–80% from pre-2020 (B)
口径警告:本页综合四份深度研究交叉整理(其中两份已于 2026-07 重跑修订:原截断源补全收束、原无引文源重写为 97 条可溯源引文;存量分级维持,新增按新源分级)。「夜经济总规模」口径已随重跑到货但极宽慎用:宽口径序列 2014 年 15.9 万亿→2020 年破 30 万亿(接近社零量级,只作氛围不作论据);窄口径「夜间旅游总花费」2024 年预计 1.91 万亿(+21.7%,B);2025 年 1–10 月游客夜间消费占全天金额 32.9%(B)——宽窄相差一个数量级,本页仍以分业态口径为主:所有数字都是分业态口径且互不可比:KTV 三套数(门店 12 万→5 万/「规模以上企业」3.99 万/文旅部「娱乐场所」6.0 万个)勿直接连线;剧本杀两套数(2020 年 60.5 亿窄口径 vs 2021–25 年 154→268→238 亿宽口径)本页采宽口径序列、窄口径入脚注。业态黄金期起止年在各文档间相差 1–2 年,时间轴按区间带处理。酒水加价 3–5 倍为区间估算(C/D);美团 vs 抖音核销率为券商研报单源(B−);海康「90%+ 准确率」为厂商口径(C)。本图未覆盖(四文档共同盲区):夜市地摊与文旅夜游(灯光秀/不夜城)、城市夜间治理制度(夜间区长/Night Mayor)、海外 nightlife——引用请注意边界。 Basis warning: cross-compiled from four deep-research reports (two re-run in Jul 2026: the truncated one completed, the uncited one rewritten with 97 verifiable citations; legacy grades stand, new figures graded on the new source). A «total night-economy size» now exists but is extra-wide — handle with tongs: the broad series runs ¥15.9T (2014) → past ¥30T (2020), near total-retail scale (atmosphere, never argument); the narrow «night-tourism spend» forecasts ¥1.91T for 2024 (+21.7%, B); night spending took 32.9% of tourists’ daily value in Jan–Oct 2025 (B) — an order of magnitude apart, so this page stays format-specific: every number is format-specific and non-comparable: KTV carries three bases (120k→50k stores / 39.9k «above-scale firms» / 60k licensed venues) that must not be joined into one line; script-murder carries two (¥6.05B narrow 2020 vs the ¥15.4→26.8→23.8B broad series) — the broad series is used here, the narrow in footnotes. Format golden-age years differ by 1–2 years across sources; the timeline uses bands. The 3–5× drink markup is an estimated range (C/D); the Meituan-vs-Douyin redemption gap is single-sourced (B−); Hikvision’s «90%+» is the vendor’s figure (C). Not covered (the sources’ shared blind spots): street night markets and light-show tourism, municipal night governance (night mayors), and overseas nightlife — mind the boundary when citing.
诚实层 · 今夜谁买单The honesty layer · who pays for tonight
欢唱价是引流品,酒水加价才是生意The karaoke price is the bait; the drink markup is the business
读图前先看三本账:①门店的账——酒水小食约占 KTV 营收 60%、加价 3–5 倍(C/D),9.9 元欢唱是获客成本不是价格;②平台的账——佣金+竞价+达人三重税,美团 2025 年推广及用户激励支出 744.6 亿(A),到店酒旅毛利率高达 87%、休闲娱乐团购综合抽佣 10%–20%(B)——全链最肥的一段在平台;③价值流向的账——利润正从场所流向平台、设备/SaaS 商与酒水供应链,场所自己几乎不做 AI,AI 是被上游「装进来」的:没有关系粘性或肉身壁垒的店,沦为重资产履约壳。Three ledgers before the map: ① the venue’s — drinks and snacks make ~60% of KTV revenue at 3–5× markup (C/D); the ¥9.9 karaoke deal is an acquisition cost, not a price; ② the platform’s — the triple tax of commission, bidding and influencers; Meituan spent ¥74.46B on promotion and incentives in 2025 (A), its in-store & travel margin runs 87%, and leisure group-buy all-in takes hit 10–20% (B) — the chain’s fattest cut sits with the platform; ③ the flow ledger — profit drains from venues to platforms, device/SaaS vendors and the liquor chain; venues barely build AI — AI is installed into them from upstream: a shop without relational stickiness or bodily moats becomes a heavy-asset fulfilment shell.
哑铃化 · 中间地带死亡带The dumbbell · the dead middle
无人端 · 冷光常亮The unmanned end · cold light on
24h 自助台球/棋牌/迷你K/电竞酒店:物联网门禁+自动计费+异常识别,成本压到极限——价值归 SaaS 与设备商24-hour self-service billiards, mahjong rooms, mini-K and e-sports hotels: IoT locks, auto-billing, anomaly detection, costs at the floor — value accrues to SaaS and device makers
中间 · 灯管熄灭The middle · tube out
普通量贩 KTV、连锁足疗、大众剧本杀——被 AI+平台+成本三面夹击,最先死The ordinary chain KTV, the mid-market foot spa, the mass-market script-murder shop — squeezed on three sides by AI, platforms and cost, first to go
人味端 · 暖光溢价The human end · warm-light premium
商务 KTV/高端洗浴/夜店 Livehouse:妈咪记得你、技师的手感、DM 的控场——反算法溢价;「无算法」正在成为高端叙事(D)The club KTV, the premium bathhouse, the livehouse: the hostess remembers you, the therapist’s hands, the DM’s command — the anti-algorithm premium; «no algorithm» is becoming the luxury story (D)
关系含量试纸:你的迎宾能不能被自助机替代?能——去左端把成本压到底;不能——去右端把人味做成溢价。站在中间,两头挨打。The relationship litmus: can a kiosk replace your greeter? If yes — go left and crush costs; if no — go right and price the warmth. Stand in the middle and both ends hit you.
Reading the MapReading the Map

从这张图带走的五条规律Five patterns to take away

立场声明:本页是批判性、祛魅的行业结构分析,用 A–D 角标区分文旅部公报/财报与厂商宣传、区间估算,口径打架处标 ⚠️ 两边并存;「无算法空间」等判断级论点已明示为 D。有偿陪侍等灰色地带仅作行业结构标记,不作展开。产品指南是市场地图,不构成经营、投资或消费建议。核心判断一句话:杀死夜间业态的是智能手机,AI 是补刀者与分选器——它决定谁能活、以什么形态活,却始终造不出「想聚」的念头。 Stance: a critical, demystifying structural analysis; A–D badges separate ministry bulletins and filings from vendor pitches and estimated ranges, ⚠️ kept where bases clash; judgment-grade claims («the algorithm-free space») are marked D. Grey zones (paid companionship) are flagged structurally, never elaborated. The product guide is a market map — not business, investment or consumption advice. The core judgment in one line: the smartphone killed the night formats and AI finishes and sorts the wounded — deciding who survives and in what form, while never once conjuring the urge to gather.