The news
OpenAI stated that users of Apple Intelligence showed little interest in the ChatGPT integration. Court documents filed in connection with ongoing proceedings indicate the company described the feature as one that “dramatically underperformed” expectations.
Context
The integration allowed Apple Intelligence users to access ChatGPT for certain requests when the on-device models were insufficient. It launched in 2024 as part of Apple’s broader push into generative AI features across iOS, iPadOS, and macOS. Prior to the launch, OpenAI had positioned the partnership as a significant distribution opportunity for its technology.
The arrangement gave Apple users an optional path to a more capable model without requiring a separate ChatGPT subscription or app switch. Apple kept the routing decision in its own hands, surfacing the ChatGPT prompt only after its on-device systems determined they could not handle the query. This design reflected Apple’s stated preference for keeping most processing local while offering an external fallback for edge cases.
Details
The 9to5Mac report, citing The Financial Times, draws directly from the court filings. OpenAI expressed disappointment with the actual usage numbers once the feature became available. No specific adoption percentages or absolute user counts appear in the available reporting, but the filings characterize the outcome as substantially below internal projections.
The integration itself was optional for users and surfaced only in specific scenarios. Apple retained control over when and how the ChatGPT option was presented. OpenAI’s comments in the filings focus on the gap between anticipated reach through Apple’s installed base and the observed engagement.
Because Apple set the trigger conditions, OpenAI had limited ability to increase visibility or change the default behavior. The company could not promote the option inside the operating system or adjust the threshold at which its model was offered. This constraint meant that even a large installed base of compatible devices did not automatically produce high call volume to OpenAI’s servers.
Reactions / counterpoints
No public statements from Apple addressing the usage figures have appeared in the court documents referenced by the reporting. The filings contain only OpenAI’s internal assessment of the results.
Why it matters
This disclosure undercuts the narrative that broad platform partnerships automatically translate into high usage for frontier AI models. Even with hundreds of millions of compatible devices, the data shows that default on-device capabilities and user habits can limit third-party model adoption. For OpenAI, the result highlights the difficulty of converting distribution deals into active usage when the partner controls the user experience.
The outcome also reveals a structural limit in how such partnerships are structured. When one company owns the operating system and the decision surface, the other company’s technology becomes a secondary option that users encounter only after the primary system has already declined the request. In practice, that positioning reduces the chance that the external model will be chosen at all, especially if the on-device alternative is fast and private.
For Apple, the low uptake aligns with its conservative routing logic. By keeping most requests on device and only offering ChatGPT when necessary, the company avoided sending large volumes of user data to a third party. The usage numbers suggest this approach matched actual demand rather than undercutting a popular feature. It also preserved Apple’s control over privacy boundaries and latency, two areas where users have historically shown strong preferences.
The episode further illustrates how litigation can surface internal performance metrics that companies would otherwise keep private. Court records from disputes between large technology firms now serve as one of the few public windows into real-world adoption data for new AI capabilities. Without those filings, OpenAI’s assessment of the Apple Intelligence integration would likely have remained internal.
The result carries implications for future deals. Any company seeking distribution through a dominant platform must now account for the possibility that the platform owner will tightly constrain when and how the partner’s technology appears. That constraint can turn a headline partnership into a low-volume channel, regardless of the size of the installed base. OpenAI’s experience supplies a concrete data point on that risk.
---
Sources:
[
{
"publisher": "9to5Mac",
"title": "OpenAI says Apple Intelligence users showed little interest in ChatGPT integration",
"url": "https://9to5mac.com/2026/09/23/openai-says-apple-intelligence-users-showed-little-interest-in-chatgpt-integration/",
"published_at": "2026-09-23T23:58:31.000Z",
"summary": "As reported by The Financial Times, new court filings show that OpenAI was disappointed by the adoption of ChatGPT through Apple Intelligence, which it says “dramatically underperformed” after launching in 2024. Here are the details. more…"
}
]
No comments yet