The head of digital strategy at a major hotel chain told me recently that her company’s direct booking rate dropped 14% in six months. The culprit was not Booking.com or Expedia. It was ChatGPT. Travelers were asking the AI assistant to plan their trips, and it was selecting hotels on their behalf, surfacing options based on the user’s history, preferences, and budget. Her brand was no longer competing against other hotels. It was competing for algorithmic permission to appear in the AI’s recommendation.
Andreessen Horowitz just published the sixth edition of its Top 100 Gen Consumer Apps report, and the data confirms what that executive felt in her P&L: AI consumer apps are no longer tools people visit. They are becoming the operating environment through which people shop, travel, manage their health, and make purchasing decisions. ChatGPT now has 900 million weekly active users, more than 10% of the global population. OpenAI reports 50 million paying subscribers. The company is testing ads, building an identity layer called "Sign in with ChatGPT," and assembling the infrastructure of a consumer super-app.
The strategic question facing business leaders is no longer whether AI matters. It is whether your company will still have a direct relationship with its customers 18 months from now.
How Three AI Platforms Are Reshaping Customer Access
The a16z report, authored by partner Olivia Moore, surfaces a divergence that has received too little attention. The three leading AI assistants are building fundamentally different ecosystems, and the choice of which ones your customers use will reshape how they find you.
ChatGPT is making the most aggressive play to become a consumer super-app. Its app directory now includes over 220 integrations across 13 categories: travel, shopping, food, health and wellness, lifestyle, entertainment. Users book flights on Expedia, order groceries through Instacart, browse homes on Zillow, and track nutrition on MyFitnessPal, all without leaving the AI interface. OpenAI has stated plainly that it wants ChatGPT to become the starting point for daily life.
Google’s Gemini is taking a different approach: embedding AI across the products billions of people already use. Gemini is now woven into Chrome, Docs, Sheets, Gmail, and Meet. In January, Google launched Personal Intelligence, connecting Gemini to Gmail, Google Photos, YouTube, and Search so the assistant can reference a user’s hotel booking, purchase history, photo library, and watch habits without being told. Gemini does not need an app store. It has the largest distribution platform in history.
Anthropic’s Claude has taken a third path entirely, building for developers and knowledge workers. Its integrations skew toward financial data terminals like PitchBook and FactSet, developer infrastructure like Snowflake and Databricks, and research platforms like PubMed. Of the combined ChatGPT and Claude app catalogs, the two share only 41 apps in common, roughly 11% of the total, almost entirely the horizontal productivity stack everyone already uses: Slack, Gmail, Google Calendar, HubSpot.
The a16z report draws an analogy to the mobile OS wars, where Apple and Google built trillion-dollar ecosystems around different philosophies. But the comparison understates the disruption. When businesses adapted to mobile, they built apps and retained their brand identity. AI platforms are different. When a consumer asks ChatGPT to find a flight to Tokyo, the AI selects which options to surface, how to frame them, and what to recommend. The airline does not control the interface. It competes for a slot in the AI’s answer.
Google controlled discovery. Amazon controlled distribution. AI platforms are positioned to control both, plus the recommendation, the transaction, and the customer relationship itself
The Biggest AI Platform Shift Is Also The Biggest Opportunity
It would be a mistake to read this data purely as a threat. The same report documents an extraordinary entrepreneurial expansion. Claude Code, Anthropic’s command-line developer agent, reached a $1 billion annualized revenue run rate in just six months. Genspark, a horizontal AI agent platform, announced $100 million in annual recurring revenue and raised a $300 million Series B. Vibe coding platforms like Lovable and Cursor are enabling people with no engineering background to build functional software. An open-source agent called OpenClaw went from a solo developer’s side project to the most-starred repository on GitHub in weeks, then was acquired by OpenAI.
The companies thriving in this environment share a common trait: they are building on the AI platform shift rather than hoping it stalls. The businesses at risk are the ones treating AI assistants as a novelty rather than a channel. Early estimates suggest that 5 to 6% of all ChatGPT usage is dedicated to commercial research. At 900 million weekly users, that is tens of millions of purchase-related sessions every week. Companies that have not mapped their presence inside these ecosystems are already invisible to a growing share of their customers.
Meanwhile, legacy software companies are proving that distribution beats invention. This is the first edition of the a16z list to include "AI-enhanced" products alongside AI-native ones: Canva, Notion, CapCut, Freepik, Grammarly. Canva now generates roughly 800 million monthly web visits, more than DeepSeek, Grok, Claude, Character, and Perplexity combined. Notion’s paid AI attach rate surged from 20% to over 50% in a single year. The incumbents did not build foundation models. They made AI invisible inside products people already depended on.
Four Questions You Should Be Asking Right Now
The a16z report tracks where usage is growing and where capital is flowing. What it does not do, by design, is ask the governance questions that growth data demands. Those questions now belong to boards and executive teams.
The first is channel dependency. If your customer discovery, product comparison, and purchasing increasingly run through an AI assistant, your business depends on a platform you do not control. The hotel executive I spoke with had no AI platform strategy six months ago. She is building one now because her revenue forced the question. Every company with consumer-facing revenue should be asking the same thing: what percentage of our customer acquisition now flows through AI, and what happens when that percentage doubles?
The second is data asymmetry. Every consumer interaction with an AI assistant generates behavioral data the platform captures and the business does not. The AI knows what the consumer searched before finding your product, what they compared it to, and whether they converted. That information advantage compounds with every session. It is the same dynamic that made Google and Amazon powerful, concentrated in a single conversational interface that controls more of the funnel.
The third is bundling risk. The report documents how standalone creative tools like Midjourney have lost ground as ChatGPT and Gemini absorbed image generation into their core products. Midjourney fell from the top 10 to number 46 in three years. This compression pattern will repeat in other verticals. If your business delivers a service that an AI platform can replicate as a built-in feature, the window for differentiation is narrowing fast.
The fourth is the agent transition. Horizontal AI agents now execute multi-step tasks on a consumer’s behalf: researching, comparing, booking, buying. Meta acquired the agent platform Manus for an estimated $2 billion. When AI agents negotiate and transact for consumers, brand loyalty is no longer a relationship between company and customer. It becomes a variable in an algorithm.
What Companies Should Do About AI Platform Strategy
The companies that navigate this transition will not be the ones with the best internal AI tools. They will be the ones that understand how AI platforms now mediate their relationship with customers.
The first move is mapping your AI presence. Find out how your products appear when consumers ask ChatGPT, Gemini, and Claude for recommendations in your category. If you do not show up, or show up unfavorably, that is the AI equivalent of being buried on page three of Google a decade ago, except the consumer never sees a page two.
The second is investing in the four moats AI cannot easily replicate: proprietary data that improves with use and creates flywheel effects; brand trust that takes years of consistent delivery to earn; human judgment and relationships that algorithms cannot substitute; and distribution that is embedded so deeply into customer workflows that switching is painful. Companies that own one of these moats have leverage. Companies that own none are commodity inputs waiting to be disintermediated.
The third is building for the AI interface. Just as companies redesigned for mobile 15 years ago, they now need to optimize for how AI assistants discover, evaluate, and recommend their products. That means structured data, clear value propositions that language models can parse, and integration with the connector ecosystems these platforms are building.
The fourth, and most important, is putting AI platform strategy on the board agenda. The a16z report makes clear that this is not a marketing problem or a technology problem. It is a strategic question about who owns the customer relationship. That question belongs in the boardroom.
The AI app charts are interesting. What they reveal about the future of customer access is existential.
Note: Originally published in Forbes.
