Facebook has long featured streams of sensational articles meant to draw clicks. Now Meta is producing its own versions of such material using artificial intelligence.

The standalone Meta AI app includes a For You section that offers a list of clickbait-style stories. All topics, images, and text are created by AI and show the typical flaws of machine-generated work.

The Meta AI app launched in April 2025 with a public Discover feed of AI images and chats from other users. That feed has been removed. The app now offers a standard chat interface along with a For You page that has existed for several months and shows suggested article prompts. Tapping a prompt creates a full story.

For a reporter in London the prompts leaned heavily on British themes such as tea, etiquette, pubs, the royal family, football, and queuing. Examples included A royal butler finally settled the milk first debate, 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 less coherent, such as When a bit of a pickle means total disaster.

A colleague received suggestions aimed at luxury watch enthusiasts, including My fake Rolex experiment and The brutal math behind the Rolex waitlist illusion.

The generated text read like padded filler that restated the prompt without adding depth. No sources were provided.

One story about a royal butler appeared to draw from a 2018 BBC Three comedy series. A Rolex experiment story seemed entirely invented as a first-person account without attribution. Other pieces cited unnamed experts or nonexistent studies.

Repeated taps on the same prompt produced similar but varied versions of the story. Entering the same headline in a new chat yielded a different result. Chat history revealed hidden system prompts that referenced internal instructions and metadata.

Each article included an image. Many were harmless cartoonish scenes, yet some showed real public figures with clear errors. One image depicted two versions of Queen Elizabeth II years after her death. Other figures resembled members of the royal family but with distorted features. Typical AI flaws such as malformed hands and unnatural poses appeared throughout. One image was an animated GIF showing impossible body movements.

It was unclear whether the app should generate images of real people under Meta policy, yet it did so. The company has stated it wants users to recognize AI content and adds labels in some cases, but no labels appeared in the feed or articles.

Meta did not answer detailed questions about the feature purpose, whether the output counts as news or fiction, existing safeguards, or compliance with its own rules on images of public figures.

A spokesperson stated the goal is to suggest what is most relevant to you such as fitness advice, meal plans, or other insights before you even have to ask. A later statement removed the word proactively. A final statement said the test was limited and would be discontinued with no plans to continue the feature.

Questions remain about how limited the test was when multiple colleagues also saw the content, what proactively meant, and why the feature was created at all.