Meta created its own AI-generated Clickbait feed

Facebook has been flooded with Clickbait posts. Now, Meta is creating its own Clickbait and AI articles.

Meta AI’s standalone app now has a “For You” section with a list of Clickbait articles for you to read. But the headlines, images, and text are all AI-generated — and that’s hardly what you’d expect from AI-generated content.

Meta AI software first he started in April 2025 with a focus on the public “Discover” which showed AI-generated images and conversations from other users (who often seemed unaware that they were being exposed). It’s all gone. The app now has a chatbot-style interactive interface, including a For You page that’s been around for a few months, which displays a list of content that prompts and, when clicked, creates a whole “story.”

Directed by me, a journalist based in London, the commentary was fiercely British, covering topics such as tea, manners, pubs, the royal family, football – sorry, football – and, naturally, the art of drawing lines. Featured articles included “Royal butler solves first milk dispute” (tea goes first, obviously), “The psychology of queuing without knowing why,” “The face of British stuttering,” and “Inside the horror game of visiting any UK pub.” Some didn’t make sense, like “When a little pickle means total disaster.”

My friend, meanwhile, seems to be placed firmly within the top watch aficionado bracket by the algorithm. His feed featured articles titled “My fake Rolex attempt” and “The cruel math behind the Rolex scam”.

AI-generated text reads like stuffy and puffy, offering little substance beyond the repetitive repetition required. There was no more money to be made.

I tried to find out where these “stories” might have started. The story of the royal tea appears to be from a series of three BBC 2018 series called Miss Hollandwhich follows a fictional beauty queen from a small Dutch town as she travels to Britain and learns “how to be noble and classy” from former royal butler Grant Harrold. The story of the “Rolex experiment”, seems to be completely fictional, created in our chat box as a first-person story without a line, after the tone that occurs when chatbots start. Some articles relied on the vague testimony of unnamed experts or fictitious research.

After tapping the same cards several times, the stories that were created were borderline fast and they were all pretty much the same thing, but slightly different. Posting the same topic in a different chat elicited a completely different response. A clear offer came from my social media profile. It showed a hidden, hidden thought that was supposed to lead to the release of the document. Others began:

“You are a conversational agent. The user is responding to the prompt card shown to them. The cards below show what prompted the user’s message,” followed by what appear to be pointers to internal instructions, information, and metadata.

The articles included pictures. Most of these were harmless – just plain pictures of people, places, and food. But others depicted real people, including public figures, and had many errors. “Who exactly pays for the royal family in 2026?” he portrayed Queen Elizabeth II, even though she had died a few years earlier and was only one person.

Next to the Queen were people who looked like other royals: Queen Kate-ish face on the left, Prince William’s unusual experiment in the background, and King Charles in the center who bears an exaggerated resemblance to his late father. Some images had normal AI expressions like impossible hands and bodies leaning at strange angles. One image became a GIF of an elderly couple dancing and making arm movements that no human body could make.

It’s not clear whether the app can create AI images of real people based on Meta, instead. lawsbut it was like that. The company has already done so he said it wants “people to know when they see AI-generated content” and where they simply add characters to other user-generated content when AI is discovered. Despite this, there was no obvious indication or sign in the food or the text that anything was AI-engineered.

Meta declined to answer many of my questions about the purpose of the session, whether the company considers published or fictional stories, what security measures are in place, and whether images of real people and celebrities fit its AI policy.

“We’re experimenting with a daily feed that shares tips, content, and ideas based on your interests,” Meta spokeswoman Tracy Clayton said in a statement. “The goal is to show what’s most important to you — like fitness advice, meal plans, or other information — before you ask.”

Clayton later posted a similar “edited” statement, ironically removing the word “prompt.”

A third statement from Clayton followed later in the day: “This was a test for a few readers and will be removed. Meta has no plans to move forward with this feature.”

This leaves me with additional questions. How small was this test if, apart from me, at least three of my three friends Seaside had a chance to use AI clickbait again? What does “proactively” mean? And, of course, who asked any of this in the first place?