核心发现 — 趋势向好
与上一周期相比,AI 提及率从 57.8% 降至 46.9%(-10.9pp),负面率从 19.7% 降至 13.6%(-6.1pp),正面率基本持平(28.5% vs 28.1%)。
关键信号:用户对 AI 的"应激反应"正在消退,评论更多开始聚焦内容本身。爆量剧 Fated to My Ruthless Alpha 负面率进一步降至 6.0%(上期 13.6%)。
但新上线剧集表现分化明显,Where Were You When I Fell 和 Billionaire's Seduce Game 正面率尚可,The Sniper King's Return 和 Emergency Reckoning 反馈偏负。
1全局指标趋势对比
46.9%
AI 提及率
↓ 10.9pp(上期 57.8%)
11.9%
配音问题提及率
↓ 2.2pp(上期 14.1%)
13.6%
负面率
↓ 6.1pp(上期 19.7%)
7.6%
希望真人演员
↓ 7.6pp(上期 15.2%)
两期全局数据对照
| 指标 | 4.27-5.06(上期) | 截至5.14(本期) | 变化 | 判断 |
| 评语总量 | 467 | 369 | -98 | 新剧上量中 |
| AI 提及率 | 57.8% | 46.9% | ↓ 10.9pp | 显著改善 |
| 配音问题 | 14.1% | 11.9% | ↓ 2.2pp | 小幅改善 |
| 表演自然度 | 7.3% | 4.1% | ↓ 3.2pp | 改善 |
| 连贯性 | 2.8% | 1.6% | ↓ 1.2pp | 改善 |
| 要真人演员 | 15.2% | 7.6% | ↓ 7.6pp | 大幅改善 |
| 正面率 | 28.1% | 28.5% | +0.4pp | 持平 |
| 负面率 | 19.7% | 13.6% | ↓ 6.1pp | 显著改善 |
2情感分布对比(正面 / 中性 / 负面)
Fated to My Ruthless Alpha
32%
62%
6%
30 Days to be the Mafia King's Bride
44%
46%
10%
Keeping His Baby and His Heart
27%
58%
16%
Billionaire's Seduce Game For Love
33%
57%
10%
Where Were You When I Fell
23%
61%
16%
Doormat No More
19%
52%
29%
The Sniper King's Return
25%
63%
13%
Love The Way You Lie
13%
75%
13%
3核心问题深度拆解
3.1 AI 提及率(46.9%,较上期 -10.9pp)
AI 提及率虽然下降,但仍有近半用户在评语中主动提到 AI。用户态度已从"一刀切拒绝"分化为三层。
"AI copy cat!!!! Acting is so fake!!!! and voice over is ridiculous, no real emotions!"
— 30 Days
"I have started to really like these wee A.I stories.. they're good yes!! However, I'd prefer real life actors more."
— Where Were You When I Fell
"This is a masterpiece, completely loved the show. Create more AI dramas"
— Billionaire's Seduce Game
变化趋势:上期以"强烈拒绝"为主,本期"有条件接受"的比例明显增加。关键差异是——用户开始承认故事本身不错("story is good""loved the show"),只是附带提到 AI 不完美。这是从"形态拒绝"到"内容评价"的健康转变。
3.2 配音/口型/口音问题(11.9%,较上期 -2.2pp)
配音问题有所改善,但在 Fated(19.0%)中仍为第一技术痛点。
"The dubbing is off. Words don't match the lips and the accents keep changing mid-sentence."
— Fated to My Ruthless Alpha
"Enough of the damn AI! It reeks of faux acting that is not consistent and undesirable dubbing!"
— The Sniper King's Return
"It was good, but sometimes the mouth didn't line up with what they were saying and it just is not the same without real actors."
— Billionaire's Seduce Game
细分问题:
• 口音切换:同一角色美式↔英式随机跳变(仍为最高频投诉)
• 口型不同步:嘴部动作和音频明显不匹配
• 语音不一致:角色声音跨集发生变化
对比上期,"发音错误"类投诉有所减少,说明专有名词发音已有修复。
3.3 连贯性 / 穿帮问题(1.6%,较上期 -1.2pp)
整体下降,但新剧 The Sniper King's Return 有严重个案。
"Gold teeth in one frame, regular teeth in the next, chopped off arm mostly, some frames two arms, dress on girl horrible... AI needs an upgrade!"
— The Sniper King's Return
"You have a mighty small wardrobe. The women wore the same outfits for days and days."
— Billionaire's Seduce Game
洞察:续存剧的连贯性投诉已接近归零,说明质检流程在起作用。但新剧(尤其 Sniper King)存在肢体/牙齿/服装等极端穿帮,说明新剧质检覆盖度不足。
3.4 表演自然度(4.1%,较上期 -3.2pp)
下降最显著的维度,说明 AI 生成的人物动作和表情已有可感知的进步。
"AI action.. way AI hardcore fake"
— Where Were You When I Fell
"I overall like the AI better than the actual actors. If there's a scene where a person is rich, you can create that in AI, but you can't create it at all in a filmed story."
— Emergency Reckoning
洞察:从 7.3% 降到 4.1%,说明用户对"假感"的抱怨在减少。甚至出现了认为 AI 在场景搭建上优于真人拍摄的声音——虽然是个例,但标志着用户认知的边界在松动。
4续存剧集趋势变化
以下为两期均有评语的剧集,可做纵向对比。
| 剧集 | 期数 | 评语数 | AI提及率 | 配音问题 | 正面率 | 负面率 |
| Fated to My Ruthless Alpha |
上期 | 191 | 44.5% | 21.5% | 33.0% | 13.6% |
| 本期 | 100 | 40.0% ↓ | 19.0% ↓ | 32.0% | 6.0% ↓ |
| 30 Days to be the Mafia King's Bride |
上期 | 160 | 66.2% | 4.4% | 25.0% | 19.4% |
| 本期 | 39 | 41.0% ↓↓ | 12.8% | 43.6% ↑ | 10.3% ↓ |
| Keeping His Heart and His Baby |
上期 | 59 | 66.1% | 18.6% | 25.4% | 25.4% |
| 本期 | 97 | 57.7% ↓ | 9.3% ↓ | 26.8% | 15.5% ↓ |
| Doormat No More |
上期 | 23 | 73.9% | 8.7% | 13.0% | 47.8% |
| 本期 | 31 | 64.5% ↓ | 9.7% | 19.4% | 29.0% ↓ |
| Love The Way You Lie |
上期 | 4 | 50.0% | 50.0% | 50.0% | 0% |
| 本期 | 16 | 31.2% ↓ | 6.2% ↓ | 12.5% | 12.5% |
趋势总结:所有续存剧的 AI 提及率和负面率均在下降。最显著的变化是:
• 30 Days:AI 提及率从 66.2% 骤降到 41.0%(-25pp),正面率从 25% 跳升到 43.6%,说明随着早期"应激用户"退出,留下来的是对内容感兴趣的用户
• Doormat No More:负面率从 47.8% 降到 29.0%,但仍是所有续存剧中最高的
• Fated to My Ruthless Alpha:负面率降到 6.0%,几乎不构成风险
5新上线剧集表现
本期新增 4 部剧集,首次有评语数据。
新剧 Where Were You When I Fell(31条)
AI 提及率为所有剧中最低(29.0%),说明 AI 痕迹控制较好。但正面率也不高(22.6%),主要问题是剧情本身("got really dumb at the ballroom")。部分用户开始"习惯"AI("I have started to really like these wee AI stories"),是正向信号。
新剧 Billionaire's Seduce Game For Love(30条)
正面率 33.3%、负面率仅 10.0%,表现接近爆量剧 Fated to My Ruthless Alpha。连贯性问题(服装重复)被提及 6.7%,需关注。有忠实 AI 用户评价为"masterpiece"。
新剧 The Sniper King's Return(16条)
AI 提及率偏高(56.2%),连贯性穿帮被具体指出("gold teeth in one frame, regular teeth in the next, chopped off arm")。样本尚小,但 AI 可见度需加强质控。
新剧 Emergency Reckoning: My Brother's Fury(5条)
仅 5 条评语,但无一条正面。AI 排斥(3/5)和内容质量("Pure garbage")同时被吐槽。需密切观察后续数据是否改善,否则有成为高风险样本的可能。
6趋势洞察
6.1 AI "脱敏"趋势确认
全局 AI 提及率从 57.8% 降至 46.9%,"希望真人演员"从 15.2% 降至 7.6%——这意味着用户的初始应激反应正在消退。尤其是续存剧,用户群体经过自然筛选后,留下来的人对 AI 的耐受度更高。
"I have started to really like these wee A.I stories.. they're good yes!! However, I'd prefer real life actors more.."
— Where Were You When I Fell
"I overall like the AI better than the actual actors. If there's a scene where a person is rich, you can create that in AI, but you can't create it at all in a filmed story."
— Emergency Reckoning
6.2 配音问题持续优化但仍是主要技术痛点
整体配音提及率下降 2.2pp(14.1%→11.9%),Fated 从 21.5% 降到 19.0%,Keeping 从 18.6% 降到 9.3%。但 Fated 仍有 19% 提及配音——口音切换和发音错误依然存在。
6.3 负面率大幅下降是最积极的信号
全局负面率从 19.7% 降到 13.6%(-6.1pp),Fated 从 13.6%→6.0%,Doormat 从 47.8%→29.0%。这说明:
• 强烈排斥 AI 的用户已经用脚投票离开(上期集中差评后不再回来)
• 留存用户的态度在软化——虽然还会提到 AI,但不再给差评
• 但正面率没有同步上升(28.1%→28.5%),说明"不排斥"≠"喜欢",内容质量仍需持续提升
6.4 新剧的"AI可见度"呈两极
Where Were You When I Fell(29.0%)和 Billionaire's Seduce(40.0%)的 AI 提及率明显低于上期平均水平,说明AI 制作质量在提升;但 The Sniper King(56.2%)和 Emergency Reckoning(60.0%)仍处于高位。新剧的技术执行水平不均匀,需建立上线前质检标准。
7结论与建议
7.1 三大核心结论
1. 用户对 AI 的排斥正在自然消退 — AI 提及率 -10.9pp,"要真人演员" -7.6pp,负面率 -6.1pp,所有核心负向指标均在改善。
2. "不排斥"尚未转化为"喜欢" — 正面率几乎未变(28.1%→28.5%),用户停留在"可以接受"阶段,还需要更好的内容来驱动正面评价。
3. 新剧制作质量参差不齐 — 最好的新剧(Where Were You)AI 提及率仅 29%,但最差的(Emergency Reckoning)高达 60%,说明制作流程标准化仍需加强。
7.2 下一步行动建议
-
继续优化 Fated to My Ruthless Alpha 的配音口音问题P0
该剧负面率已降至 6%,但仍有 19% 提及配音。修复口音一致性可将其从"可接受"提升为"标杆",成为后续新剧的品质参照。
-
建立新剧上线前 AI 可见度质检标准P0
新剧间的 AI 提及率差异巨大(29%~60%)。建议制定质检清单(口音一致性、穿帮检查、口型匹配),确保上线前达到 Where Were You 的水平。
-
把 30 Days 的变化作为运营案例研究P1
该剧 AI 提及率从 66%→41%,正面率从 25%→44%,是"用户态度软化"的最佳案例。分析其留存用户画像,用于指导内容推荐策略。
-
关注 Emergency Reckoning,若后续数据不改善考虑下架P1
5 条评语 0 正面,AI 提及 60%。如果后续 20 条评语仍无改善信号,应评估止损。
-
探索"轻AI"类型方向P2
Where Were You 的数据说明,当 AI 痕迹足够轻时,用户的评价焦点会自然转向剧情——这正是 AI 仿真人剧的理想状态。可总结其制作参数作为后续制作标准。