Meta Launches AI-Driven Clickbait News Feed

Facebook has been a hotspot for clickbait articles for ages, and now, Meta is diving into the fray with its own AI-generated clickbait content. The Meta AI app has introduced a “For You” section that curates a list of sensational stories, all crafted by AI. The results are as dubious as you might expect from machine-generated content.

Initially launched in April 2025, the Meta AI app featured a public “Discover” feed showcasing AI-generated images and user conversations, often without users realizing they were public. That setup has since vanished, replaced by a standard chatbot interface and a For You page that has been around for a few months, offering a stream of suggested article prompts that, when selected, generate entire “stories.”

When I tested the app as a London-based reporter, the prompts were heavily themed around British culture, covering topics like tea, etiquette, pubs, royalty, football, and, of course, the fine art of queuing. Suggested stories included titles like “A royal butler finally settled the milk first debate” (it turns out tea goes in first), “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 prompts were even more nonsensical, such as “When a bit of a pickle means total disaster.”

In contrast, a colleague of mine seemed to be categorized as a luxury watch enthusiast by the algorithm, receiving suggestions like “My fake Rolex experiment” and “The brutal math behind the Rolex waitlist illusion.”

The AI-generated text felt like fluff, lacking depth and merely reiterating the premise of each prompt without any credible sourcing. I attempted to trace the origins of these “stories.” The royal butler tea piece appeared to reference a 2018 BBC Three comedy series called Miss Holland, which follows a fictional beauty queen learning to be posh from a real former royal butler. The “Rolex experiment” story, however, seemed entirely fabricated, generated in our chat as a first-person narrative without any attribution, following the usual whirring sound of a chatbot at work. Other stories relied on vague mentions of unnamed experts or fictional studies.

When I clicked on the same prompts multiple times, the generated stories remained within the general theme but were clearly variations of the same concept. However, entering the same headline in a different chat produced a completely different response. My chat history revealed hidden prompts meant to trigger article generation, starting with a line that read: “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 looked like internal instructions and metadata.

The articles came with images, many of which were harmless, featuring generic cartoonish characters, landscapes, and food. However, some included real people, including public figures, and were riddled with inaccuracies. One article titled “Who really pays for the royal family in 2026?” bizarrely featured two Queen Elizabeth IIs, despite her passing years ago. Surrounding the Queen duplicates were poorly rendered approximations of other royals, with a vaguely Princess Kate-like face, a strange version of Prince William, and an exaggerated depiction of King Charles. Other images displayed typical AI flaws, such as distorted hands and bodies in unnatural poses. One image even turned out to be a GIF of an older couple dancing in ways no human could replicate.

It was unclear whether the app was allowed to generate AI images of real people under Meta’s vague guidelines, but it did. The company has previously stated its intention for users to recognize AI-generated posts, claiming to label some user-generated content when AI is detected. Yet, there was no clear indication or label in the feed or articles to show that any material was AI-generated.

Meta was tight-lipped about many questions regarding the feature’s purpose, whether the output is considered news or fiction, what safeguards are in place, and whether the use of real people’s images aligns with its AI-content policies. A spokesperson mentioned they were testing a daily feed to share tailored tips and recommendations, aiming to provide relevant content like fitness advice and meal plans before users even ask.

Later, the spokesperson sent a nearly identical statement, curiously omitting the word “proactively.” A third statement followed, indicating that this was a test for a limited number of users and that the feature would be discontinued, with no plans to continue it.

This leaves me with lingering questions. How was this test limited if at least three of my colleagues also had access to the same AI clickbait feature? What did “proactively” even mean? And, most importantly, who actually requested any of this?