‘I Like My Big Rat Wife’: Meet the People Using Chatbots to Write Custom Fiction

A University of Washington analysis of more than 500,000 public ChatGPT prompts found that over one-third involved fiction generation, although usage was concentrated among heavy users and OpenAI/NBER data put fiction-related prompts at 1.4%. Researchers estimate 7% of chatbot users generate fiction even after adjusting for power users, signaling a committed niche demand for customizable AI prose, role-play, and erotica. AI fiction could expand into a tiered publishing market as model quality improves, despite current weaknesses in long-form narrative coherence and literary style.

Analysis

The investable signal is not AI-authored publishing revenue; it is evidence that high-frequency, personalized entertainment can become a durable inference workload. Role-play and iterative fiction are unusually sticky because each session accumulates user-specific context, raising switching costs and making engagement more valuable than generic search queries. For GOOG, a modest consumer-fiction use case matters chiefly if it improves Gemini retention and creates a path to premium memory, character, or creator-tool tiers rather than merely adding low-margin token volume.

The near-term economics are ambiguous. Open-source models running locally and specialized character platforms can absorb much of this demand without paying Google, while safety constraints may leave proprietary models structurally disadvantaged in adult-oriented or copyrighted-franchise use cases—the segments with the highest engagement intensity. That creates a second-order opportunity for device vendors and inference infrastructure if on-device generation grows, but it also means headline usage statistics should not be extrapolated directly into Alphabet revenue.

Over 1-3 months, watch for Gemini product telemetry: paid conversion, session duration, memory/context features, and developer/API usage rather than aggregate chatbot prompts. Over 6-18 months, the upside case is that interactive fiction becomes a proving ground for agentic consumer experiences—persistent characters, game NPCs, and personalized video—expanding the addressable market for Google Cloud inference and Android-native AI. The thesis is falsified if consumer AI engagement rises while inference costs outpace monetization, or if copyright and youth-safety enforcement sharply restricts character-based products.

Consensus may overvalue polished-content displacement and undervalue substitution away from passive media toward infinite, user-directed experiences. The likely first casualty is not premium human-authored books but low-cost digital content categories whose value proposition is formulaic abundance: mobile-game narrative content, fan-fiction platforms, and commoditized erotica subscriptions. This is a longer-duration engagement shift, not a material near-term earnings revision for GOOG.

Key Decisions for Investors

• Maintain, do not add aggressively, to GOOG on this signal alone. Treat Gemini paid-user conversion and Cloud AI inference margin commentary at the next two earnings prints as the gating catalysts; absent disclosed monetization, the revenue impact is immaterial relative to Alphabet scale.
• Set an alert for evidence of Gemini persistent-memory or character/creator subscription launches. A credible premium tier with disclosed retention or ARPU would support a 6-18 month GOOG overweight because personalized context can improve consumer switching costs more than undifferentiated chatbot access.
• Watch a relative-value basket: long GOOG versus a short exposure to commoditized digital-content incumbents only after measurable engagement displacement emerges. Candidate hedge proxies include RBLX or MTCH, but no position is warranted without platform-level time-spent and payer data; current linkage is thematic rather than earnings-visible.
• Risk-manage any GOOG AI premium exposure around regulatory and cost disclosures: reduce if management indicates rising AI capex/depreciation without Cloud margin expansion, or if copyright/safety policy materially limits consumer character interactions. Those outcomes would make engagement growth a cost center rather than a monetizable product moat.