Meta made its own AI-generated clickbait news feed
Meta made its own AI-generated clickbait news feed
Meta 打造了自己的 AI 生成式“标题党”新闻流
Facebook has long been filled with feeds of clickbait articles. Now, Meta is making its own clickbait articles with AI. 长期以来，Facebook 上充斥着各种“标题党”文章。现在，Meta 正在利用人工智能亲自炮制这类文章。
The standalone Meta AI app now has a “For You” section that populates a list of clickbait-style stories for you to read. But the topics, images, and text are all AI-generated — and as questionable as you’d expect from AI-created works. 独立的 Meta AI 应用现在增加了一个“为你推荐”（For You）板块，其中填充了一系列供用户阅读的“标题党”风格故事。然而，这些主题、图片和文本全部由 AI 生成——其质量正如你对 AI 作品所预期的那样，令人质疑。
The Meta AI app first launched in April 2025 with its focus on a public “Discover” feed that showed AI-generated images and conversations from other users (who frequently seemed unaware that they were being made public). That’s all disappeared. The app now has a standard chatbot interface, plus a For You page that’s been present for at least a few months, displaying a stream of suggested article prompts that, when tapped, generate entire “stories.” Meta AI 应用于 2025 年 4 月首次发布，当时侧重于一个公共的“发现”（Discover）信息流，展示其他用户生成的 AI 图像和对话（这些用户往往并不知道自己的内容被公开了）。现在这一切都消失了。该应用目前采用标准的聊天机器人界面，外加一个已存在至少几个月的“为你推荐”页面，展示一系列建议的文章提示；点击这些提示，即可生成完整的“故事”。
When targeting me, a reporter based in London, the prompts were aggressively British, involving topics like tea, manners, pubs, royals, football — sorry, soccer — and, naturally, the art of queuing. Suggested stories included “A royal butler finally settled the milk first debate” (the tea goes first, apparently), “The psychology of joining a queue without knowing why,” “The anatomy of the devastating British tut,” and “Inside the extreme sport of visiting every UK pub.” Some made even less sense, like “When a bit of a pickle means total disaster.” 当针对我这位驻伦敦的记者时，这些提示充满了浓厚的英伦色彩，涉及茶、礼仪、酒吧、皇室、足球（抱歉，是美式足球）以及排队的艺术等话题。建议的故事包括《皇室管家终于平息了“先加奶还是先加茶”的争论》（显然是先加茶）、《在不明所以的情况下加入排队的心理学》、《令人崩溃的英式“啧啧”声解剖》以及《探秘打卡全英酒吧的极限运动》。有些标题甚至更离谱，比如《当“陷入困境”意味着彻底的灾难》。
My colleague, meanwhile, appears to have been placed firmly within the luxury watch aficionado bracket by the algorithm. His feed suggested stories called “My fake Rolex experiment” and “The brutal math behind the Rolex waitlist illusion.” 与此同时，我的同事似乎被算法牢牢地归入了奢侈手表爱好者的行列。他的信息流推荐了诸如《我的假劳力士实验》和《劳力士等待名单幻象背后的残酷数学》等故事。
The AI-generated text read like puffy filler, offering little substance beyond repeatedly restating the premise of the prompt. Sourcing was also nonexistent. 这些 AI 生成的文本读起来就像空洞的填充物，除了反复重申提示的前提外，几乎没有提供任何实质性内容。文章也完全没有来源标注。
I tried to track down where these “stories” may have originated. The royal butler tea story appears to trace back to a 2018 BBC Three comedy series called Miss Holland, which follows a fictional beauty queen from a small Dutch town as she travels to Britain and learns “how to be posh and classy” from real former royal butler Grant Harrold. The “Rolex experiment” story, meanwhile, appeared to be a complete fabrication, generated in our chat box as a first-person narrative without a byline, after a bit of usual whirring that happens when a chatbot is generating. Other stories leaned on vague references to unnamed experts or fictional research. 我试图追溯这些“故事”的来源。关于皇室管家泡茶的故事似乎源自 2018 年 BBC Three 的一部名为《荷兰小姐》（Miss Holland）的喜剧系列，讲述了一位来自荷兰小镇的虚构选美皇后前往英国，并向真正的前皇室管家格兰特·哈罗德（Grant Harrold）学习“如何变得时髦和优雅”的故事。而“劳力士实验”的故事则似乎完全是虚构的，它在我们的聊天框中以第一人称叙述的形式生成，没有署名，且伴随着聊天机器人生成内容时常见的“嗡嗡”处理过程。其他故事则依赖于对未具名专家或虚构研究的模糊引用。
When I tapped the same cards more than once, the generated stories stayed within the rough bounds of the prompt and all were clearly versions of the same thing, but slightly different. Typing the same headline into a separate chat produced a completely different response. The clearest giveaway came from my chat history. It showed the hidden, suggested prompts that were supposed to trigger the generation of articles. One began: “You are a helpful conversational assistant. The user is responding to a proactive feed card that was shown to them. The card context below provides background on what prompted the user’s message,” followed by what appeared to be references to internal instructions, information, and metadata. 当我多次点击同一张卡片时，生成的故事都保持在提示的大致范围内，显然是同一内容的略微不同版本。在单独的聊天中输入相同的标题，则会产生完全不同的回复。最明显的破绽来自我的聊天记录。它显示了本应触发文章生成的隐藏建议提示。其中一个开头写道：“你是一个乐于助人的对话助手。用户正在响应向他们展示的主动推送卡片。下方的卡片上下文提供了触发用户消息的背景信息，”随后似乎是针对内部指令、信息和元数据的引用。
The articles had images attached. A lot of these were harmless — bland mush of cartoony people, landscapes, and food. But some depicted real people, including public figures, and were riddled with errors. “Who really pays for the royal family in 2026?” featured two Queen Elizabeth IIs, despite her death several years prior and her existence as only one person. 这些文章附带了图片。其中许多是无害的——比如卡通人物、风景和食物的平庸堆砌。但有些图片描绘了真实人物，包括公众人物，且错误百出。《谁在 2026 年真正为皇室买单？》一文中竟然出现了两位伊丽莎白二世女王，尽管她已于几年前去世，且历史上只有一位女王。
Around the Queen clones were people who seemed to be approximations of other royals: a vaguely Princess Kate-ish face to the left, a strange attempt at Prince William at the back, and a sort-of King Charles in the middle who bore an exaggerated resemblance to his late father. Other images had usual AI tells like impossible hands and bodies leaning at unnatural angles. One image actually turned out to be a GIF of an older couple dancing and making arm movements no human body could make. 在女王克隆体周围，是一些看起来像是其他皇室成员的近似形象：左侧是一张模糊的凯特王妃脸庞，后方是对威廉王子的奇怪尝试，中间则是一位看起来像查尔斯国王的人，但他长得极其夸张地像他已故的父亲。其他图片则有常见的 AI 破绽，比如不可能出现的手部姿势和以不自然角度倾斜的身体。其中一张图片甚至是一个 GIF 动图，显示一对老年夫妇在跳舞，并做出了人类身体无法完成的手臂动作。
It wasn’t clear whether the app should be able to generate AI images of real people in accordance with Meta’s own, rather opaque rules, but it was. The company has previously said it wants “people to know when they see posts that have been made with AI” and that it automatically adds labels to some user-generated content when AI is detected. Despite this, there was no obvious indication or label in the feed or articles that any material was AI-generated. 目前尚不清楚该应用是否应该根据 Meta 自身相当模糊的规则生成真实人物的 AI 图像，但它确实做到了。该公司此前曾表示，希望“让人们在看到 AI 制作的帖子时能够知晓”，并称在检测到 AI 内容时会自动为部分用户生成的内容添加标签。尽管如此，在信息流或文章中，没有任何明显的迹象或标签表明这些材料是由 AI 生成的。
Meta declined to answer many of my questions about the feature’s purpose, whether the company considers the output news or fiction, what safeguards are in place, and whether images of real people and public figures comply with its own AI-content policies. Meta 拒绝回答我关于该功能目的的许多问题，包括公司是否将这些输出视为新闻或虚构内容、采取了哪些保障措施，以及真实人物和公众人物的图像是否符合其自身的 AI 内容政策。
“We’re testing a daily feed that proactively shares tips, content, and recommendations tailored to your interests,” Meta spokesperson Tracy Clayton said in a brief statement. “The goal is to suggest what’s most relevant to you – such as fitness advice, meal plans, or other insights – before you even have to ask.” “我们正在测试一种每日信息流，主动分享根据您的兴趣量身定制的提示、内容和建议，”Meta 发言人特雷西·克莱顿（Tracy Clayton）在一份简短声明中表示。“其目标是在您提出要求之前，就为您提供最相关的内容——例如健身建议、膳食计划或其他见解。”
Clayton later sent a nearly identical “updated” statement, mysteriously removing the word “proactively.” 克莱顿随后发送了一份几乎完全相同的“更新”声明，神秘地删除了“主动”（proactively）一词。
A third statement from Clayton followed later in the day: “This was a test for a limited number of users and it will be deprecated. Meta has no plans to move forward with this feature.” 当天晚些时候，克莱顿发出了第三份声明：“这是一项针对少数用户的测试，目前将被弃用。Meta 没有计划继续推进此功能。”