I Tried ChatGPT in CarPlay and Immediately Started Hallucinating.

Back in February, it was revealed that Apple planned to integrate voice assistants into CarPlay . This news came with the first beta of iOS 26.4, although Apple didn’t actually provide testers with the opportunity to test the integration in the software. Instead, the company confirmed the news in its CarPlay developer guide , stating that support for the “CarPlay voice app” would require at least iOS 26.4.
Well, last week Apple released iOS 26.4 , and while the update brought a number of new features, none of them involved AI-powered chatbot integration in CarPlay. The company even released the first beta of iOS 26.5 without mentioning the upcoming feature. I admit, the idea of using ChatGPT, Claude, or Gemini with CarPlay completely disappeared from my mind, replaced by thoughts of when Apple’s major AI-powered Siri update would finally arrive. But OpenAI hasn’t forgotten.
The ChatGPT app is now available on CarPlay.
In a surprising move, OpenAI added ChatGPT support to CarPlay. The company announced the news on Thursday in a post on X , accompanied by the following tagline: “The voice experience you know is now available on the road.” In the post, OpenAI confirmed that the feature works on iPhones with the latest version of iOS (iOS 26.4) and, of course, in cars that support CarPlay.
I happen to have CarPlay in my car, so I decided to give it a try. My iPhone already has iOS 26.4, so it wasn’t a problem, but I did have to update the ChatGPT app to the latest version. Before, I didn’t see the app in CarPlay. After the update, ChatGPT appeared on the last page of apps in the CarPlay window.
I tested it, and it works pretty much as expected: when you launch ChatGPT in CarPlay, it opens directly in voice mode. (After all, you’re driving.) There are just two buttons on the screen: “End,” which ends the session, and “Mute,” which mutes the audio, presumably so you can talk to other people in the car (or yell at other drivers?) without fear of ChatGPT eavesdropping.
Once you launch the app, you can simply talk to ChatGPT, and it will respond. It’s a truly driver-friendly version of the familiar voice mode.
CarPlay doesn’t stop ChatGPT from hallucinating.
I have nothing else of significance other than this: in the very short time I was testing this feature, the AI almost immediately started hallucinating. I asked if it understood what I was doing, and it said no. When I pressed it, thinking it might figure it out given the context, it admitted it knew I was using CarPlay, so I was probably driving. I asked how long this feature had been around, and it mistakenly thought I was asking about CarPlay, so it replied, “Since 2014.” When I corrected that and asked how long ChatGPT integration had been in CarPlay, it replied, “Late 2023.”
“The end of 2023?” I asked. “Yes, exactly!” ChatGPT replied. “So it’s still fairly recent, but it’s expanded significantly since then, especially with updates from OpenAI and Apple. So, yes, you could say you’re at the very beginning of this wave!” I asked if it knew what year it was. ChatGPT acknowledged that it was 2026, meaning “several years have passed since the initial launch,” and thanked me for noting that.
“Okay,” I said. “But I was under the impression this feature was brand new, literally just released today.” This was the first time in the entire conversation that ChatGPT had to “think about it,” as hinted by the faint melody playing in the background. When the app responded, it confirmed the feature had launched in iOS 26.4. “This is the first time third-party chat apps like ChatGPT have been available in CarPlay, so this is a brand new feature!” When I asked why it had initially listed 2023, it said it was a mistake. When I asked for the source of this information, it said it “probably accidentally remembered an earlier timeline.”
This isn’t a new problem. All generative AI systems, not just ChatGPT, are prone to hallucinations , and that’s to be expected when using this technology. But I can’t help but think that CarPlay only exacerbates the problem: you can’t easily fact-check the AI while driving; all you can do is ask it to fact-check itself. Sure, it might correct the error, as it did when I challenged it, but it might also simply generate a hallucination with the correction. If you rely on this tool to answer your questions while driving, you might simply receive incorrect information that you mistake for the truth.
Note: In April, Lifehacker’s parent company, Ziff Davis, filed a lawsuit against OpenAI, alleging that it infringed Ziff Davis’s copyright in the training and operation of its artificial intelligence systems.