Putting ChatGPT inside hundreds of millions of iPhones sounds like a slam dunk. It wasn’t. According to court documents from OpenAI’s antitrust battle with Elon Musk’s SpaceXAI, the ChatGPT integration with Siri was so underused that OpenAI’s own internal teams called it “dramatically underperforming” within weeks of launch. That’s a brutal self-own for a deal that was supposed to reshape how people interact with AI on mobile.
The details came out because OpenAI needed them to. The company revealed the embarrassing numbers in newly unsealed filings specifically to shoot down Musk’s claim that the Apple partnership was an illegal exclusive deal that blocked Grok from Siri and strangled competition. To prove the deal wasn’t a monopoly, OpenAI had to prove it was basically a flop. And that’s exactly what they did.
The integration went live in December 2024 as part of Apple Intelligence, and things started going wrong almost immediately. The core problem was friction. Apple tucked the ChatGPT opt-in behind a multi-step setup process buried in the Settings app. Most users never found it. And even the ones who triggered Apple’s Visual Intelligence feature, which can analyze what your camera sees, mostly ignored the ChatGPT prompt or just opened the standalone ChatGPT app directly from their home screen. Why jump through hoops when the icon is right there?
By January 2025, barely a month after rollout, OpenAI had already revised its user growth forecast downward. But the disappointment went beyond low sign-ups. Apple’s strict privacy terms meant OpenAI couldn’t collect query data or use Siri traffic to train its models. So there was no subscriber bump and no training data. The commercial value of the deal essentially evaporated. By March 2026, the situation had gotten tense enough to spark a round of heavily redacted conversations between the two companies.
Meanwhile, Apple quietly started building its Apple Foundation Models using Google’s Gemini Frontier models for distillation. That shift left OpenAI without the two-year exclusivity window it had originally asked for. So not only did the integration underperform, but Apple was already looking elsewhere.
Now here’s where it gets interesting from a legal angle. Musk’s lawsuit claimed the OpenAI-Apple deal was an illegal exclusive arrangement that locked Grok out of Siri, stunted X’s growth into a so-called super app, and even prevented Apple from being pressured to drop its 30% App Store cut. OpenAI brought in economic expert Dr. Catherine Tucker to dismantle that argument. Her finding: the share of AI users actually accessing ChatGPT through Apple Intelligence was, in her words, virtually unmeasurable. OpenAI also pointed out the deal was explicitly non-exclusive from day one. Apple reserved the right to add other AI providers and, notably, never actually did. Musk eventually dropped his claims against Apple and left OpenAI as the only defendant heading into a trial scheduled for January 2027.
This story matters beyond the courtroom drama. It’s a real-world stress test of what happens when a top AI product gets plugged into a top mobile platform and the results are… nothing much. The lesson isn’t that AI integrations are doomed. It’s that friction kills adoption fast, and burying a feature in settings is basically the same as not shipping it. For Apple, it’s also a reminder that being the phone in everyone’s pocket doesn’t automatically make your AI features sticky. Users have choices, and they’ll take the path of least resistance every time. Apple is going to need to do a lot better if it wants Apple Intelligence to actually mean something to the people carrying its hardware every day.