Meta is experimenting with AI chatbots that have personalities to convey to the company’s four billion users around the world.

Examples of the “personas,” as Meta calls them, are a surfer giving weather updates (“those are some gnarly-lookin’ clouds, dude”) or another that supposedly speaks like Abraham Lincoln.

The practical purpose is the usual one—to use the chatbots to answer questions and make recommendations—but they also will entertain users and, Meta hopes, make people more likely to choose Meta for their searches as competition among social media platforms becomes even more intense.

The larger purpose is to collect more data on more users, enabling the company to build personal profiles of individuals and then use AI to tailor ads to their specific interests.

“Once users interact with a chatbot, it exposes much more of their data to the company, so that the company can do anything they want with that data,” AI researcher and ethics advisor Ravit Dotan told the Financial Times. 

Having all that data tempts companies into “manipulating and nudging” users, she added.

The company will unveil more details about the project at its “Connect” developers conference next month, Meta chief Mark Zuckerberg said in a 2 August earnings call. Meta is developing AIs for a variety of customer-oriented and back-office tasks, he added, and could create AIs to animate avatars in the metaverse.

Meta has made earlier forays into AI, but those chatbots slipped too easily into passing along misinformation or fell into hate speech. Now the company is working on filters that could screen out inappropriate questions and run a check on a chatbot’s output before it blurts incorrect or offensive things.

The idea of chatbots with personalities is not new. Character.ai generates chats in the speaking style of Tesla boss Elon Musk or video game character Mario.