Apple’s Siri is hitting some serious roadblocks. The tech giant’s AI assistant lags behind competitors, with human testers rating its performance as merely adequate. Recent attempts to upgrade Siri have stalled indefinitely, while rivals like Google and Meta surge ahead. Despite introducing new features at WWDC25, including an AI Workout Buddy and enhanced Image Playground, Apple’s struggles continue. Former executives point to persistent bugs and ineffective improvement strategies. The full story behind Apple’s AI challenges reveals an even more concerning picture.

Nearly all of Apple’s AI efforts have hit a technological brick wall. The tech giant’s newly announced AI models are falling short of competitors, with human testers giving their “Apple On-Device” model merely a polite nod of adequacy when compared to similar offerings from Google and Alibaba.
Even worse, their beefier “Apple Server” model can’t keep up with OpenAI’s year-old GPT-4, and Meta’s Llama 4 Scout is showing it up in image analysis. The model’s capability to process 3 billion parameters reflects its limited problem-solving potential.
Apple’s server AI stumbles behind GPT-4 and Meta’s Llama 4 Scout, falling short in both general performance and image processing capabilities.
The situation with Siri is particularly painful. That promised upgrade? Indefinitely delayed. It turns out cramming large language models into a decade-old voice assistant isn’t exactly a walk in the park. The new Live Translation feature in Messages, FaceTime, and Phone calls offers some hope for improved communication capabilities.
Former executives are spilling the tea, revealing that every attempt to modernize Siri leads to a fresh batch of bugs. Their strategy of gradual improvements rather than a complete overhaul is looking about as effective as putting racing stripes on a horse and buggy.
The timing couldn’t be worse. While competitors are sprinting ahead in the AI race, Apple’s research division is stuck trying to figure out how to make Siri understand the difference between “set a timer for pasta” and “set a timer for my pasta.”
Some frustrated customers have even lawyered up, suing Apple for marketing AI features that are still stuck in the digital waiting room.
There’s a glimmer of hope, though. At WWDC25, Apple managed to roll out some new AI tricks – developer access to on-device foundation models, enhanced Image Playground with ChatGPT integration, and an AI Workout Buddy for Apple Watch.
They’re even introducing something called “Liquid Glass” interface. But let’s be real – these are just side dishes when everyone’s waiting for the main course: a Siri that actually works as promised.
The clock is ticking. While Apple’s SVP of Software Engineering, Craig Federighi, points to quality standards as the reason for delays, the real question is whether Apple can evolve Siri before users completely lose patience.
Right now, it’s looking like a classic case of too little, too late.