Character.AI

Character.AI remains the defining consumer franchise in AI entertainment: more than 20 million monthly active users on a platform whose median engaged session times still dwarf ordinary consumer apps, a character-creation flywheel with millions of user-generated personas, and a balance sheet transformed by the August 2024 Google transaction — roughly $2.7B paid to license its models non-exclusively and rehire Noam Shazeer and Daniel De Freitas, cashing out early investors around a $2.5B mark while leaving the company independent and funded. The 2025-2026 chapter under CEO Karandeep Anand (ex-Meta, Instagram commerce) is a deliberate repositioning from "AI companion" to "AI entertainment platform": AvatarFX animated video characters, Scenes interactive storytelling, the Imagine creative suite, a discovery feed — multimodal, creator-driven formats that monetize through c.ai+ subscriptions (subscriber growth reported around 250% in the first half of 2025) plus a nascent ads business placed at navigation moments rather than inside conversations. Even the painful decision — ending open-ended chat for under-18s in November 2025, ahead of legal compulsion — reads as strategic strength: it cut roughly 10% of users to de-risk the other 90%, reassured advertisers, defused the worst litigation optics, and set a compliance standard smaller rivals will struggle to match. The asset that remains is scarce and real: the largest dedicated audience in interactive AI entertainment, category-defining brand recognition, and network effects in user-created content. If interactive synthetic media becomes a mainstream entertainment format — the trajectory every engagement metric supports — Character.AI is its incumbent.

Character.AI is a company that sold its brain and kept the audience. The Google deal extracted Shazeer, De Freitas, and the core research team; the models that remain are, by the company's own admission, increasingly fine-tuned open-weight models rather than frontier internal research — which means its conversational quality now rides the same commodity curve available to every competitor. What differentiated it is gone; what remains is engagement under siege from every direction: wrongful-death and product-liability lawsuits (the Setzer and Peralta cases most prominently) that will run for years, an FTC companion-chatbot inquiry, state statutes like California's SB 243, and the self-imposed under-18 ban that removed a disproportionately engaged cohort and clipped the growth curve. Monetization remains immature relative to the ~$2.5B mark investors were bought out at — subscriptions off a small base, ads barely launched — while inference costs on tens of millions of users burn cash, and August 2025 reporting had the company exploring a sale or new funding as costs piled up, hardly the posture of a business compounding from strength. Competition now comes from distribution monsters: xAI ships companions inside Grok, Meta embeds AI characters across three-billion-user surfaces, OpenAI is loosening adult-content restrictions, and Chai out-executes on efficiency below. The pivot to "entertainment platform" is a crowded, capital-intensive fight (video generation against Sora-class models) undertaken with a weakened research bench. The plausible terminal states — absorbed for its audience, or slowly commoditized as character chat becomes a feature — both price well below the legend.