Where does consumer AI stand at the end of 2025?

Market Consolidation Trends

• The general LLM assistance space is trending toward a "winner-take-most" model, with only 9% of consumers paying for more than one service from the major group (ChatGPT, Gemini, Claude, Cursor).
• Less than 10% of ChatGPT users visited other major LLM providers (like Gemini) for most of the year.
• ChatGPT maintains a dominant lead with 800–900 million weekly active users, while all other major competitors (Claude, Grok, Perplexity) trail with 8–10% of usage volume.
• Growth rates are diverging significantly: Gemini's desktop users grew 155% year-over-year (accelerating at scale), whereas ChatGPT grew 23% year-over-year.
• Google holds a distribution advantage on mobile, with Gemini at 50% of ChatGPT's scale on Android compared to only 17% on iOS.

2025 Product Launches and Capabilities

• Multimodal Breakthroughs: Image and video models drove the year's viral trends; OpenAI led with the "Ghibli moment" (ChatGPT 4.0 image) and Sora 2, while Google led with VO3, VO3.1, and the viral NanoBanana/NanoBanana Pro models.
• Reasoning and Realism: Image models have advanced beyond aesthetics to include complex reasoning, physics simulation (e.g., correct background movement), and multi-step tasks (e.g., updating monopoly board properties with specific names).
• Search Integration: Realism in image generation now relies heavily on search integration to ensure accuracy in tasks like product photography and historically accurate image generation.
• Interface Strategies: OpenAI generally keeps features (Pulse, group chats, shopping) within the core ChatGPT interface, whereas Google frequently launches standalone products (Google AI Studio, Labs) to allow for custom interfaces.
• Perplexity Comet Browser: Launches saw sustained high traffic for Perplexity's browser, outperforming ChatGPT's own browser (Atlas) in sustained usage despite lower initial distribution.

Strategic Challenges and Disagreements

• Social Features: OpenAI's social initiatives (Group Chats, Sora 2) are viewed as bearish for native social app success because they lack the "status game" required for viral participation; Sora 2 currently functions more as a creator tool than a consumption platform.
• Anthropic's Positioning: Claude is highly rated by technical users for "opinionated" workflows and artifacts but struggles to reach the mainstream consumer due to steep usability barriers and a lack of intuitive onboarding.
• Product Execution Gap: Big labs (OpenAI, Google) are structured for incremental improvements and risk-averse promotion, often resulting in failed standalone consumer UI experiments (e.g., ChatGPT Pulse, Google Gems), whereas startups succeed by building "opinionated" products.
• Compute Constraints: Labs face a strategic tension between allocating compute resources to training new models versus inference for viral consumer features (e.g., releasing NanoBanana might delay the next major LLM training run), a constraint not present for app-layer startups.

Forward-Looking Statements for 2026

• Consumer Builder Boom: Experts predict 2026 will be a "huge year" for consumer builders as models have reached a quality threshold sufficient for scalable, standalone applications.
• Multimodal Convergence: Future models will shift from "text in, image out" to "anything in, anything out," merging text, image, and video reasoning into unified "mega models."
• Enterprise-to-Consumer Spillover: High growth in ChatGPT Enterprise (8–9x YoY) may drive future consumer adoption as users transition to using the platform for work and personal life simultaneously.
• App Generation Evolution: While big labs will attempt app generation, startups are expected to lead in "opinionated" app interfaces due to the labs' structural inability to build niche, risk-taking products.
• Monetization Shifts: Consumer AI is seeing revenue retention rates over 100% for the first time, driven by usage-based pricing (buying extra tokens) in addition to subscriptions.

Recommended Tools and Workflows

• Creative Tools: Creai is highlighted for managing character consistency and style templates across models; Level Labs Reader is recommended for converting reading material to audio.
• Productivity & Design: Gamma is recommended for text-to-slide generation with flexible sizing; Granola for context-aware note-taking; Perplexity Comet browser for agentic task execution.
• Experimental AI: Pameli (Google Labs) is cited as a key example of combining agents with image generation for ad campaign creation; GPT-5.2 in Cursor is recommended even for non-coders for knowledge work and essay writing.