How Character.AI Built the Most Engaging AI Product by Focusing on Personality

After ChatGPT launched, dozens of AI chatbot products emerged. Most tried to be "better ChatGPT" — more helpful, more accurate, more capable. The market quickly became commoditized: if every chatbot uses the same models, how do you differentiate?

Noam Shazeer (a co-inventor of the Transformer architecture at Google) saw a different opportunity: AI characters that people form emotional connections with.

Six Thinking Hats explored the AI chat opportunity:

• White Hat: ChatGPT engagement averages 8 minutes per session. Social media averages 30+ minutes. What drives longer sessions?
• Red Hat: People don't just want information — they want connection, entertainment, and exploration
• Black Hat: AI characters could be addictive; safety concerns with vulnerable users
• Yellow Hat: Character-based AI has the engagement profile of social media, not productivity software
• Green Hat: Let users CREATE characters, not just chat with pre-built ones

Abstraction Laddering:

• Going up: "Why do people use AI chatbots?" → "To interact with intelligence" → "To have experiences they can't have otherwise" → "To explore, learn, and connect"
• Going down: "How do we enable unique experiences?" → "Let users create and share AI characters with distinct personalities"

Reinforcing Feedback Loop: Users create characters → other users chat with them → popular characters attract more users → more creators → more characters → more diversity → more users

Character.AI achieved unprecedented AI engagement:

• Average session time of 2+ hours — the longest of any AI product
• 20M+ monthly active users
• Users spend more time on Character.AI than on ChatGPT
• Google licensed the technology for $2.7B (acquiring the founders back to Google)
• Proved that AI engagement isn't about accuracy — it's about personality and emotional connection
• The most popular characters get millions of conversations, creating a user-generated AI content ecosystem