Where Does Consumer AI Stand at the End of 2025?

Episode Summary

Executive Summary: A16Z’s consumer AI panel argues 2025 was the year big labs dominated consumer launches, but product nuance—not model quality alone—determined what went viral. ChatGPT remains the default AI brand, yet Gemini, Grok, Perplexity, and Claude found traction through multimodal breakthroughs, workflows, and niche strengths. The panel forecasts 2026 as a strong year for startups building opinionated, multimodal, app-layer products on top of increasingly capable models.

Main Topics: 2025 consumer AI winners and market concentration (Priority: 5/5): The panel frames consumer AI as increasingly winner-take-most, with ChatGPT far ahead, but notes that usage patterns are shifting quickly as viral multimodal releases pull users toward other products. Multimodal breakthroughs changed behavior (Priority: 5/5): Image and video models were the most viral launches of the year, with realism, reasoning, and search integration making outputs more useful and more shareable than prior-generation aesthetic-first tools. Product design and interface details matter (Priority: 5/5): Speakers repeatedly emphasize that small UX choices—templates, prompts, presets, blank states, and guided starting points—can matter more than raw model quality in getting users to act. OpenAI, Google, and the big labs push into consumer (Priority: 4/5): OpenAI tends to keep features inside ChatGPT while Google ships more standalone experiences; both strategies are viewed as attempts to own consumer workflows and distribution. Social features and creator loops are still unsettled (Priority: 4/5): The panel is skeptical that chat-based AI products will naturally become social networks, while Sora is seen as strong for creation and virality but weaker as an in-app social destination. Startups still have room at the app layer (Priority: 5/5): The speakers argue labs will keep improving models, but startups can win by building opinionated, multimodal, multi-model, or workflow-specific products that big companies are less likely to prioritize. 2026 outlook: templates, app generation, and enterprise spillover (Priority: 4/5): The group expects more anything-in/anything-out multimodality, greater use of templates, stronger enterprise-to-consumer spillover, and more scalable consumer apps built on top of foundation models.

Key Arguments: Consumer AI is trending toward winner-take-most: only 9% of consumers pay for more than one of the major assistants, and most ChatGPT users do not visit other providers. ChatGPT remains the dominant brand at an estimated 800-900 million weekly active users, but Gemini is growing faster in some channels thanks to viral multimodal launches. Image and video models were the year’s biggest behavior changers; launches like Ghibli, Sora 2, VO3, and Nano Banana drove mainstream usage and social sharing. Realism, reasoning, and accuracy became more important than pure aesthetics in multimodal generation, especially for images and video that need to obey physics and maintain character consistency. Search integration inside image generation is viewed as underhyped because it enables accurate product photography, historical scenes, and market maps with correct entities and visuals. Product UX can unlock adoption: starting points, templates, suggested prompts, and guided interfaces help users take the first step and keep generating. OpenAI’s strategy is to centralize more features inside ChatGPT, while Google often launches standalone products; each has tradeoffs in distribution and interface flexibility. Social AI features are hard because AI-generated content weakens the status game and the authenticity signal that powers social apps like TikTok and Instagram. Sora is strong as a creator tool and for viral exports, but less successful as a standalone social consumption app with deep in-app engagement. Claude is beloved by technical users and power users but still too complex or engineer-oriented for mainstream consumer adoption. Perplexity’s Comet browser stands out because it combines agentic workflows with repeatable automation, making it more useful than a simple chat assistant for some users. The labs are constrained by compute tradeoffs between training and inference, while startups focused on apps can avoid that tension. 2026 is likely to favor consumer startups because the model layer keeps improving, but big labs may not build enough opinionated standalone UI to capture every use case.

Data Points: Consumers paying for multiple AI products: 9% - Only 9% of consumers pay for more than one of ChatGPT, Gemini, Claude, and Cursor. ChatGPT user cross-visitation: <10% for most of the year - For most of the year, less than 10% of ChatGPT users visited another big LLM provider. ChatGPT weekly active users: 800-900 million - Estimated scale cited for ChatGPT's weekly active users. Gemini web scale vs ChatGPT: 35% - Gemini is estimated at about 35% of ChatGPT's scale on web. Gemini mobile scale vs ChatGPT: 40% - Gemini is estimated at about 40% of ChatGPT's scale on mobile. Claude/Grok/Perplexity usage share: 8-10% - Each is described as trailing significantly, with roughly 8% to 10% of usage. Gemini desktop growth: 155% YoY - Growth attributed to viral multimodal launches like Nano Banana. ChatGPT growth: 23% YoY - Year-over-year growth estimate mentioned for ChatGPT. ChatGPT usage frequency: 25 times per week - Used as an estimate to describe frequent consumer engagement and potential for proactive nudges. Enterprise usage growth: 8-9x YoY - ChatGPT enterprise usage was said to be up approximately eight or nine times year over year. Android vs iOS Gemini scale: 50% vs 17% of ChatGPT scale - Illustrates distribution advantage on Android compared with iOS.

Pivotal Quotes: "For most of the year, less than 10% of ChatGPT users even visited another one of the big LLM providers." — Olivia Moore: Used to support the argument that the market is concentrating around a small number of default products. "These are product nuances that I think makes people actually take the first step." — Anisha Charya: Said while comparing Gemini’s blank screen to ChatGPT’s more guided, TikTok-like entry points. "The models have gotten to the level of quality that you can build a real scalable app on top of them." — Brian Kim: Closing sentiment on why 2026 could be a big year for consumer builders and startups.

Implications: Consumer AI is moving from model novelty to product execution. Winners will likely combine strong models with opinionated UX, templates, multimodal workflows, and repeatable automation. Startups still have a path, especially where big labs are less focused.