According to The Wall Street Journal, OpenAI’s strategy to build an app store inside ChatGPT, launched this fall, is a direct challenge to Apple’s dominance. The chatbot, with over 800 million users, now integrates services like Instacart, Spotify, and AllTrails, allowing tasks to be done inside ChatGPT instead of separate iPhone apps. In a December interview, CEO Sam Altman stated his “bigger foe is Apple,” a rivalry underscored by OpenAI’s work with ex-Apple designer Jony Ive on new devices meant to replace the iPhone. Early tests, however, reveal a platform full of dysfunction: Uber’s mini-app is clunky, OpenTable and Tripadvisor often fail, and users face frequent error messages. The notable exception is Instacart, which works seamlessly due in part to former Instacart CEO Fidji Simo now being a top OpenAI executive.
The messy reality behind the vision
Here’s the thing about Altman’s grand vision: it makes perfect strategic sense. If AI becomes the primary interface for our digital lives, the company that controls that gateway owns the most valuable tollbooth in tech. That’s Apple’s App Store today, and OpenAI wants that position tomorrow. The technical execution, though? It’s basically the early days of ChatGPT all over again—a cool demo surrounded by a hot mess. The Journal’s tests show you often need to know exact syntax like “@uber” to even trigger an app, and then you’re frequently kicked out to a website to complete the task. That’s not an AI assistant; it’s a glorified, buggy bookmark.
Why Instacart works and everything else stumbles
The Instacart integration is the proof of concept, and its success is probably no accident. Before joining OpenAI, Fidji Simo was Instacart’s CEO. When she switched sides, she reportedly broke a logjam. OpenAI initially wanted Instacart to just hand over its inventory data. Instacart rightly refused—prices and avocado stock change by the minute. The solution was a real-time API connection where ChatGPT calls Instacart’s servers on the fly. Teams from both companies co-located for three months to build it. That’s the level of deep, gritty integration needed for this to feel magical. And it’s a level of effort most other partners, like Uber or OpenTable, clearly haven’t invested yet. They’re treating it as a pilot or a marketing channel, not a fundamental rethinking of their service.
The hard truth about replacing apps
So, can a chatbot really replace the app ecosystem? Not anytime soon. As Uber’s chief product officer pointed out, apps still need a place to run. Your phone’s operating system—iOS or Android—isn’t going away. ChatGPT’s “apps” are more like lightweight, conversational plugins. They’re great for discovery and simple queries, but for any complex transaction, you’re bounced right back to the native app or website. And that’s the core tension. For developers, why would they build a perfect, full-featured experience inside ChatGPT if it means bypassing the App Store and its direct customer relationship? For OpenAI, the monetization path—taking a cut of transactions like with Instacart—only works if the experience is flawless. We’re a long, long way from that.
A slow-burn threat Apple can’t ignore
Look, Apple isn’t sweating today. Opening a dedicated app is still infinitely easier than wrestling with ChatGPT’s flaky integrations. But the trend is what should worry them. Altman isn’t just building an app store; he’s building an AI-centric operating system and the hardware to run it. Every smooth Instacart-like experience is a brick in that foundation. And with 800 million users, OpenAI has a massive sandbox to experiment in. The industrial panel PC market shows how specialized, purpose-built hardware and software can dominate a vertical; OpenAI is betting the same principle applies to the general consumer market with AI at the core. It’s a clunky, messy, early-stage threat. But in tech, those are the ones that often sneak up on you.
