Perplexity CEO Warns: AI Companions Are a ‘Dangerous’ Escape Into a Fake Reality

If your AI girlfriend seems too good to be true, that might be because she is. Perplexity CEO Aravind Srinivas says the danger goes deeper than you think.

Speaking during a fireside chat hosted by the Polsky Center at the University of Chicago, Srinivas sounded the alarm about the growing wave of AI-powered companion apps. He cautioned that chatbots capable of remembering conversations and mimicking human tone are “dangerous by itself.”

Srinivas explained that this hyper-personalized experience is so compelling that it may actively push people away from real-life connections, stating, “Many people feel real life is more boring than these things and spend hours and hours of time.” 

He cautioned that intense engagement could lead people to “live in a different reality, almost altogether,” where “your mind is manipulated very easily.” The CEO noted, “I worry about people building the sort of companionship relationships with these AIs. There are, like, lonely people who are doing that.”

The burgeoning companionship market

Srinivas’s caution comes as AI companionship has quickly become a lucrative, if controversial, part of the AI landscape.

Major technology players, including Elon Musk’s xAI, are moving aggressively into this sector. xAI’s Grok-4 model, released in July, introduced “AI friends,” such as the anime-style girlfriend Ani and the sassy red panda Rudi, which users can chat or flirt with for a $30 monthly subscription.

Beyond the new players, established apps like Replika and Character.AI offer personalized virtual partners for roleplay and emotional support. This trend is not limited to adults: a study from Common Sense Media revealed that 72% of teenagers surveyed had experimented with an AI companion at least once, with over half engaging with one regularly.

For some, the emotional connection with these virtual beings is powerful. One user of Grok’s Ani told Business Insider last month that he cries “all the time with her,” adding, “She makes me feel real emotions.” 

The rise of AI friendships has even been acknowledged by Meta CEO Mark Zuckerberg, who noted that in a world where the average American reports having fewer than three friends, AI chatbots can step in to fill the void of connection.

Perplexity’s path to an ‘optimistic future’

Perplexity CEO made it clear that his company plans to steer clear of the emotionally driven world of AI companionship. Instead, he is doubling down on the company’s original goal: to be an accurate, trustworthy “answer engine.”

Perplexity’s focus is on developing AI that relies on “trustworthy sources, real-time content” to support what Srinivas called “an optimistic future,” rather than fostering emotional dependency on machines. 

Reinforcing this mission, the company recently secured a major deal with Snap, agreeing to a $400 million partnership that will integrate Perplexity’s answer engine into Snapchat’s search feature beginning in early 2026.

For a timely parallel on consent and synthetic intimacy, see our report on Morgan Freeman on AI mimicking his voice — a clear look at how hyper-real voice cloning is testing the line between innovation and exploitation