大约一周前拿到一份电子表格后,我的脑袋一直在嗡嗡作响。
在深入晾晒我的脏衣服之前,先说一个显而易见的事实:数百万人隐约感觉到 AI 正在改变一切。而 LLM 确实能快速给出答案,非常方便。我和我的团队每天都在使用 Claude 和其他工具。
但真正亲身经历过这场颠覆的速度与烈度的人,要少得多。不是一年后,不是六个月后,而是此时此刻。
所以,让我用自己的书作为桌上的尸体,向你展示一场死亡是什么样子。
首先,一些更宏观的统计数据
2026 年第一季度,Publishers Weekly 报道称,“成人非虚构类”较 2025 年第一季度下降了 9%。谁知道呢……也许这符合历史波动?
但仔细看,自助类降幅最大,销量同比下降 26.3%。 16 个子类别中只有两个——手工艺/爱好/古董/游戏和宗教——实现了增长(分别为 9.6% 和 1.6%)。仅这些例外本身就足以写一篇有趣的博客文章,留待日后再说。
但说实话:一个季度并不能构成趋势。
所以让我们拉远视角,看看我几年来的完整目录。
我的个人销售数据
以下是我的五本书——《每周工作4小时》、《4小时身体》、《4小时厨师》、《巨人的工具》和《导师部落》——作为一个组合的国内纸质书销量数据(BookScan)。
请记住,这些书全都是《纽约时报》和/或《华尔街日报》的畅销书第一名,而且《每周工作4小时》在出版整整十年后的2017年,仍是全亚马逊被划线标注最多的书籍之一。销量一直出人意料地持久……而且可预测。这些书长期以来一直是我可以依靠的年金。
但可惜!
天堂里出了麻烦:
| 年份 | 同比 |
| 2022 | 基准 |
| 2023 | -5% |
| 2024 | -13% |
| 2025 | -46% |
| 2026(按运行率计算) | 较 2025 年下降 57% |
让这一点先沉淀一下。
ChatGPT 由更新后的 GPT-3.5 模型驱动,于 2022 年 11 月 30 日发布。
2023 年轻微下滑 -5%,然后 2024 年 -13%,接着地板消失了:2025 年 -46%,随后今年更是加速到 -57%。如果这一运行速率持续下去,我的书目在 2026 年的纸质书销量将比 2022 年大约减少 80%,而其中几乎所有下滑都发生在 Claude 和 ChatGPT 等 LLM 的使用爆发之后。
但电子书和有声书呢?
看 2025 年该书目所有格式(纸质书 + 电子书 + 有声书),下半年相比上半年下降了约 45%。当然,这里面也有各种保留意见。
我们可以谈论 Amazon 的库存变化、疫情后支出转移、少数潜在例外、异常事件后的均值回归(例如 2024 年《The 4-Hour Body》因Gary Brecka而走红),等等。
但即使我尽力为反方论点做最强辩护……这些也都只是花言巧语和一厢情愿。我不相信任何一堆脚注能开始解释指导性非虚构类作品近乎垂直的下跌。
世界上许多最强的自助类品牌——那些长期占据主导地位的耀眼宠儿——也正遭受重创。这些还是最顶尖的表现者。在所有卖书的地方,你都能在端架上看到它们。但如果你看看 2025 年与迄今为止 2026 年的 BookScan 销量,再做点计算,情况并不好看。如果你合理地假设自助类书籍往往在上半年销量最高,那么我能想到的那些最大牌的作者,销量会下滑约 40–60%。
我的经纪人手头有几十年的报表可供对比,他说得很直白:2025 年是第一次大幅下滑,2026 年看起来更严重,而在这段时间里唯一真正发生变化的事情,就是 AI 的加速发展。
一些出版商把原因指向 YouTube 和播客的增长,这些确实有影响,但我认为它们相对而言只是可以忽略的零头。
到底发生了什么?
从功能上讲,想想我的书到底是什么。
在某种程度上,The 4-Hour Body就是一张查询表。我把自己的很多书都描述成《选择你自己的冒险》式的菜单:我该怎么减脂?我该怎么改善睡眠?我该怎么快速增肌 10 磅?同样,The 4-Hour Workweek是一棵用于设计你的生活方式、让你的收入自动化的决策树。
2019 年,获取这些答案的最佳界面是一本书。
到了 2026 年,数百万人相信,最佳界面是一个免费的聊天机器人,它读过我的书——以及成千上万本别的书——能在 15 秒内给你一份个性化方案,并根据你的体重、你的日程、你的伤病,以及你对茅屋奶酪的厌恶做出调整。
那么,我能分享一些有力的反方论点吗?可以,我也会这么做,但这一趋势只会加速并加剧。这一大趋势几乎不会放过建议行业中的任何人。
指导性非虚构类作品是煤矿里的金丝雀吗?
如果“操作指南”类书籍正被碾压,因为 LLM 似乎能提供更快、更便宜、更个性化的建议……下一个被砍掉的会是什么?或者说,还有什么容易被 AI 生成的替代品取代?
- 操作指南类 YouTube 视频。当你需要的只是其中 40 秒的内容时,为什么要拖着进度条翻遍一段 24 分钟的视频?AI 可以替你观看,然后把步骤直接交到你手上。
- 指导性播客。播客收听中很大一部分——包括我的节目的很多内容——就是在对话中挖掘可付诸行动的建议。如果 AI 能从 800 多期节目中提取、总结并个性化这些要点,还有多少人会按下播放键?AI 替代品或摘要会提供你偏好的任何格式:文本、音频、视频,或接下来出现的任何形式。根据我看过的技术演示,《头号玩家》(也许去掉触觉反馈)比人们想象的近得多。
- 在线课程、newsletter、建议类博客。同样的逻辑。任何核心价值主张是“把指令从我的脑袋转移到你的脑袋”的东西,如今都在与一个能即时、对话式且免费完成这件事的界面竞争。
我的立场——而我真心希望自己是错的——是没错,规定性非虚构类书籍就是煤矿里的金丝雀,而这座煤矿大得惊人。我相信大语言模型会成为通往一切的界面:搜索和购物,显而易见,还有刷视频、总结播客、浏览课程,甚至翻阅书籍。原始内容并不会真正消失;它只是变成了大多数人永远不会直接接触的原材料。
这对依赖广告的搜索意味着什么?这对依赖广告和订阅者的新闻业意味着什么?
当99%经过严格事实核查的媒体都藏在付费墙后面时,会发生什么?简短的回答是:人们跳过它,直接问AI。
根据Pew Research的数据,83%的美国人在过去一年中没有以任何形式为新闻付费。而当他们撞上付费墙时呢?只有区区1%的人会掏出信用卡。我用过各种工具来绕过付费墙,因为我不想新增100个订阅,但在收入的军备竞赛中,那些工具正被出版商的新工具击败。那结果怎样?我让大语言模型给我总结链接的文章,它们做得非常出色。翻译过程中丢失了很多东西,但对于快速了解动态来说已经足够好了。
还有什么能大致以当前形态存活下来?
大概有。体验,那种不仅仅是信息的东西:喜剧、娱乐、讲故事、虚构小说等等。你不会让AI去总结一场单口喜剧专场,而一部伟大小说的梗概也不是一部伟大小说。声音、品味和个性或许最终会成为唯一持久的护城河。但“给我X的5个步骤”?那是一门艰难的生意,而且即将变得艰难得多。
那为什么我并不恐慌?
坦白说:我内心有一部分觉得这反而让事情更清晰了。
我从来不是因为单位经济模型才投身写作的。我写作,是因为一本书是我所知的痴迷密度最高的传递方式——一个人两三年的生命,被压缩成你能捧在手里的东西。书在我写书之前很久就改变了我的人生,而对我最重要的那些书,从来都不是超级畅销书。
我之前承诺过要给出反方论点,那么这里就是一个:
至少对我的书而言,秘诀在于排序——事物的逻辑顺序——再加上那些极其私人的故事(例如,原宿时刻收录于《4小时身体》),它们才真正促使人们改变长期养成的习惯。
ChatGPT 的病毒式传播在 2022 年底席卷全球。但远在那之前,2010 年,《4小时身体》首次出版。全书厚达 608 页,并登上了《纽约时报》畅销书榜单。最终有数百万人买了它,但我的许多聪明朋友给我发来的消息大致都是这个意思:“我爱你,但我太忙了,读不了那本大部头。你能不能给我发几条快速要点,告诉我怎么减掉20磅?”
他们中有些人是糖尿病前期,有些人即将结婚,或者有其他理由认真对待这件事。有多少人真的照我发去的要点去做了呢?
恰好是零——他们中没有一个人——采纳了那些建议。相比之下,成千上万的读者——被引导着走过一条精心设计的路径——在终生尝试其他减肥法均告失败之后,减掉了100多磅(在此查看一些前后对比照片)。
为什么?精心规划的旅程和来自真实人物的真实故事中,依然蕴含着大量的魔力。
所以这就是我目前的落脚点,至少眼下是这样:
我宁愿为 10,000 个人写书,也不愿为 10,000,000 个人制作短视频片段。
再补充一点,我想说:
我宁愿为 10,000 个真正因书而改变的人写书,也不愿为 10,000,000 个几天或几分钟内就把它们忘掉的人批量生产短视频。
为什么?
其一,我的不少播客视频片段获得了 5000 万到 1 亿次观看,或者 5 万个点赞,或者随便你挑什么虚荣指标,但你猜这些数字如何转化为完整剧集的下载量——而重要的细微之处恰恰在那里?精确地说是零。你根本无法在图表上看到任何影响。平台越来越擅长把用户困在自己的平台上,而追逐算法是一场逐底竞争。
其二,信息的市场正在坍缩进聊天机器人。而转化的市场——即与一个为之倾注心血的头脑长时间共处一个主题——可能只会变得更小、更古怪、也更有趣。我敢打赌会是这样。某种程度上,我们正在回归互联网的早期时代。
规定性非虚构书籍的死亡——至少作为一种大众市场的信息生意——已近在眼前。当然,会有暂时的异类,但趋势线指向一个方向。
对每一位写作者、播客主和创作者来说,问题不是界面转移是否会降临到你的格式上。而是当它降临时你打算怎么做。一如既往,我认为趁还没干涸先挖好自己的井是明智之举。
那么,当脚下的沙子在不断移动时,你该怎么做?也许答案是回归基本功:
1) 找到你的1000 个铁杆粉丝。如果你一开始做得很好但后来走了弯路,是时候重新审视了。非常清楚地弄明白这 1000 个人是谁。
2) 给他们惊喜和愉悦。一次又一次地超额交付。
3) 成功了!
真的能这么简单吗?
而且这真的会有那么难吗?反向拉扯的离岸流强得离谱——算法追逐、制造标题党的激励、机器人辅助的“互动”,等等等等。AI 个性化会让这些塞壬之歌的诱惑力再强 100 倍。
但归根结底——真的还能有别的选择吗?
我把自己绑在长文的桅杆上,但也许我只是在妄想,或者嗑自己产出的东西嗑嗨了。
只有时间能给出答案,而那个时间很快就会到来。
***
同意?不同意?还是有不同角度?请在这里的评论区告诉我。
My head has been spinning after getting a spreadsheet roughly a week ago.
Before we dive into my dirty laundry, let’s state the obvious: millions of people have a vague sense that AI is changing things. And LLMs sure are convenient for getting answers quickly. My team and I use Claude and other tools daily.
But far fewer people have first-hand experience with the speed and intensity of disruption that’s happening. Not in a year, not in six months, but right now.
So let me show you, using my own books as the cadaver on the table, what a fatality looks like.
First, some broader stats
For the first three months of 2026, Publishers Weekly reported that “adult nonfiction” was down 9% from Q1 2025. Who knows… maybe in line with historical fluctuations?
But looking more closely, Self-help had the steepest subcategory decline, with units down 26.3% year-over-year. Only two of 16 subcategories—crafts/hobbies/antiques/games and religion—grew at all (9.6% and 1.6%, respectively). The exceptions alone could make an interesting blog post for another time.
But, let’s be honest: one quarter doesn’t make a trend.
So let’s zoom out and look at my full catalog over a few years.
My personal sales numbers
Below are the domestic print numbers (BookScan) for my five books—The 4-Hour Workweek, The 4-Hour Body, The 4-Hour Chef, Tools of Titans, and Tribe of Mentors—as a portfolio.
Keep in mind that all of these were #1 NYT and/or WSJ bestsellers, and The 4-Hour Workweek was one of the most highlighted books across all of Amazon in 2017, a full decade after publication. The sales have been surprisingly durable… and predictable. These books have long been an annuity that I could count on.
But alas!
There’s trouble in paradise:
| Year | Year-Over-Year |
| 2022 | baseline |
| 2023 | -5% |
| 2024 | -13% |
| 2025 | -46% |
| 2026 (run-rate) | -57% vs. 2025 |
Let that sink in for a minute.
ChatGPT, powered by the updated GPT-3.5 model, launched on November 30, 2022.
There was a gentle -5% slip in 2023, then -13% in 2024, and then the floor disappears: -46% in 2025, followed by an even steeper -57% pace this year. If the run-rate holds, my catalog will sell roughly 80% fewer print copies in 2026 than it did in 2022, with almost all of that happening since LLMs like Claude and ChatGPT exploded in use.
But what about ebooks and audio?
Looking at all formats (print + ebook + audio) for the catalog in 2025, the second half of the year was down ~45% versus the first half. Now, there are caveats, of course.
We could talk about Amazon stocking changes, post-pandemic shifts of spending, a few potential exceptions, reversion to the mean after outlier events (e.g., TikTok virality of The 4-Hour Body in 2024, thanks to Gary Brecka), and so on.
But, even if I try my best to steelman a counter-argument… it’s all fancy-talk and wishful thinking. I don’t believe any constellation of footnotes begins to explain a near-vertical drop in prescriptive nonfiction.
Many of the strongest self-help franchises on the planet—standout darlings with perennial dominance—are also getting hammered. These are the best performers. You see them on endcaps everywhere books are sold. But if you look at BookScan sales for 2025 vs 2026 thus far, and do a little math, it ain’t pretty. If you rightly assume that self-help books tend to sell the most copies in H1, the biggest names I could think of will be down ~40–60%.
My agent, who has decades of statements to compare against, put it bluntly: 2025 was the first big drop, 2026 looks more severe, and the only thing that’s really changed in that timeframe is the acceleration of AI.
Some publishers point to the growth of YouTube and podcasts, and those certainly contribute, but I think they are relative rounding errors.
What’s actually going on?
Think about what my books are, functionally speaking.
On some level, The 4-Hour Body is a lookup table. I have described a lot of my books as Choose Your Own Adventure-style menus: How do I lose fat? How do I fix my sleep? How do I quickly add 10 pounds of muscle? Similarly, The 4-Hour Workweek is a decision tree for designing your lifestyle and automating your income.
In 2019, the best interface to those answers was a book.
In 2026, millions believe that the best interface is a free chatbot that has read my books—and thousands of others—that will give you a personalized protocol in 15 seconds, adjusted for your bodyweight, your schedule, your injuries, and your aversion to cottage cheese.
Now, can I share some compelling counter-arguments? Yes, and I will, but the trend is only going to accelerate and intensify. The broad trend will spare (next to) no one in the advice business.
Is prescriptive nonfiction the canary in the coal mine?
If “how-to” books are getting crushed because LLMs seem to provide faster, cheaper, and more personalized advice… What’s next on the chopping block? Or, what is vulnerable to being replaced by AI-generated alternatives?
- How-to YouTube videos. Why scrub through a 24-minute video to find the 40 seconds you need, when an AI can watch it for you and hand you the steps?
- Prescriptive podcasts. A huge portion of podcast listening—including a lot of my show—is mining conversations for actionable advice. If an AI can extract, summarize, and personalize the takeaways from 800+ episodes, how many people still press play? The AI alternatives, or summaries, will provide whatever format you prefer: text, audio, video, or whatever comes next. Based on technology that I’ve seen demo’d, Ready Player One (maybe minus the haptics) is a lot closer than people think.
- Online courses, newsletters, advice blogs. Same logic. Anything with a core value proposition of “transferring instructions from my head to yours” is now competing with an interface that does it instantly, conversationally, and for free.
My position—and I’d genuinely love to be wrong—is yes, prescriptive nonfiction is the canary in the coal mine, and the coal mine is enormous. I believe LLMs become the interface to everything: search and purchasing, obviously, but also surfing video, summarizing podcasts, navigating courses, even browsing books. The original content doesn’t exactly disappear; it just becomes raw material that most people never touch directly.
What does this mean for search that depends on ads? What does this mean for journalism that depends on ads and subscribers?
What happens when 99% of the rigorously fact-checked media is behind a paywall? The short answer: people skip it and ask the AI.
Per Pew Research, 83% of Americans haven’t paid for news in any form in the past year. And when they slam into a paywall? A mere 1% pull out a credit card. I have used various tools to get around paywalls, as I don’t want to have 100 new subscriptions, but in the revenue arms race, those tools are getting beaten by new publisher tools. So what happens? I prompt LLMs to give me a summary of the linked articles, and they do it beautifully. There’s a lot lost in the translation, but it’s good enough for a quick update.
Will anything survive in roughly its current form?
Probably. Experience that isn’t solely information: comedy, entertainment, storytelling, fiction, etc. You don’t ask an AI to summarize a stand-up special, and a synopsis of a great novel is not a great novel. Voice, taste, and personality may end up being the only durable moats. But “give me the 5 steps to X”? That’s a tough business that’s about to get a lot tougher.
So why am I not panicking?
A confession: part of me finds this clarifying.
I never got into writing because of unit economics. I got into it because a book is the highest-density transfer of obsession I know—two or three years of someone’s life, compressed into something you can hold. Books changed my life long before I wrote one, and the books that mattered most to me were never huge bestsellers.
I promised counter-arguments earlier, so here is one:
For my books, at least, the secret sauce is in the sequencing—the logical ordering of things—plus the deeply personal stories (e.g., The Harajuku Moment in The 4-Hour Body) that actually catalyze people to change long-standing habits.
The viral sensation of ChatGPT took the world by storm in late 2022. But well before that, in 2010, The 4-Hour Body was first published. It clocked in at 608 pages and hit #1 on the New York Times bestseller list. Millions ultimately bought it, but a lot of my smart friends texted me some version of this: “I love you, but I’m too busy to read that monster. Could you send me a handful of quick bullets for losing 20 pounds?”
Some of them were pre-diabetic, about to get married, or had some other reason to take this seriously. How many acted upon the bullet points I sent?
Precisely zero—none of them—implemented the advice. In contrast, thousands of readers—who were led along a carefully designed path—lost 100+ pounds (see some before-and-after photos here) after failing other diets their entire lives.
Why? There’s still plenty of magic in meticulously planned journeys and real stories from real people.
So here’s where I’ve landed, at least for now:
I’d rather write books for 10,000 people than make short-form video clips for 10,000,000.
Adding a little more, I’d say:
I’d rather write books for 10,000 people who are genuinely changed by them than crank out short-form videos for 10 million people who forget about them within days or minutes.
Why?
For one thing, quite a few of my podcast video clips have gotten 50–100 million views, or 50k likes, or choose-your-vanity-metric, but guess how that’s translated to downloads of the full episodes, where the important nuance is? Precisely zero. You literally cannot see the impact on a graph. The platforms are increasingly better at keeping users captive on their platforms, and algorithm chasing is a race to the bottom.
Second, the market for information is collapsing into the chatbot. The market for transformation—for sitting with one mind, at length, on a subject it has bled for—might just get smaller, weirder, and more interesting. I’d bet on it. In a way, we’re reverting to the earlier days of the Internet.
The death of prescriptive nonfiction books, at least as a mass-market information business, is nigh. Sure, there will be temporary outliers, but the trend line points in one direction.
The question for every writer, podcaster, and creator isn’t whether the interface shift comes for your format. It’s what you’ll do once it does. As always, I think it’s good to try and dig your wells before you’re dry.
So how do you do that while the sands are shifting under your feet? Perhaps it’s a return to basics:
1) Find your 1,000 True Fans. If you started off doing this well but have meandered, it’s time to revisit. Get very clear on who those 1,000 people are.
2) Surprise and delight them. Overdeliver again and again.
3) Success!
Could it really be that simple?
And could it really be that hard? The riptides pulling in the opposite direction are absurdly strong—algo chasing, incentives to create clickbait, bot-assisted “engagement,” and more. AI personalization will make these siren songs 100x more seductive.
But in the end – Could there really be any other choice?
I’m tying myself to the mast of long-form, but maybe I’m delusional or otherwise high on my own supply.
Only time will tell, and that time is coming soon.
***
Agree? Disagree? Different angle? Please let me know in the comments here.