Siri isn’t an assistant, it’s an embarrassment

It’s time to call it: Siri sucks. And if we’re being honest, it’s been terrible for years.

Apple’s voice assistant was so far ahead of the competition that it had every opportunity to basically become the default for a generation. It could have been a brand that encompasses all things of its type, like how all cola is “Coke,” all facial tissue is “Kleenex,” and every personal watercraft is a “Jet Ski.” It could have been a verb, the way every web search is now “Googling it.” (The only reason Google’s assistant hasn’t stepped into that role is that the company can’t settle on a brand name.)

Now, the company is telling us that the big new Siri features that might have made Siri pretty good have been delayed for a year (maybe more). In the meantime, we don’t see any reason to believe that Siri won’t continue its pitiful pace of “improvement” that has left it miles behind its AI competitors.

Apple should be deeply embarrassed by Siri. It’s integral to every Apple device, but it’s not a good experience, and it’s so far behind in the AI race it leads one to wonder if Apple is ever going to catch up.

Missing the AI boat

When asked about AI, Apple will be quick to remind you that, really, it’s a pioneer. They’ve been using AI and machine learning throughout their products for years, ever iPhone since the iPhone X brought dedicated AI acceleration hardware.

That’s all true, but it’s also fair to say the company had no idea at all that generative AI would be so popular or powerful. And so, from image and video generation to text generation and chatbots, Apple is years behind what they’re doing at OpenAI, Google, or Meta.

Apple

Ignoring that Apple’s own text generation is mediocre, summaries of notifications and emails are often hilariously askew, and the image generation looks like MidJourney circa 2022, the “other guys” are even doing better voice assistants than Apple. I can have a full-on, free-flowing conversation with ChatGPT or Gemini.

I can have them give longer or shorter answers, ask them to change their voice or personality, and more. If ChatGPT had access to my phone functions, it would be game over. With Siri, I have to call upon years of playing Infocom text adventures to phrase my queries in just the right way to make it work like it should. Open mailbox. Go south. Open window.

That sort of thing wasn’t even on tap for the Siri update in iOS 18, the one we thought we would get in a month or two but has now been kicked down the road to next year. Conversational Siri? Once rumored to be coming in an update to iOS 19 (read: early 2026), Mark Gurman now says it’s probably going to be bumped back to iSO 20 in 2027!

In other words, Apple is still at least a year and probably two away from delivering the experience its competitors are today. By the time Apple catches up, those competitors will have advanced even further.

Apple is safe, but only because the iPhone’s core functions and the frameworks that enable access within apps are under Apple’s lock and key. Competitor’s superior AI can only work on the iPhone or iPad as far as Apple allows, so I won’t ever get the ability to ask ChatGPT or Gemini to set an alarm. I won’t ever get a hands-free, always-listening voice assistant that isn’t Siri, no matter how far behind it is.

But Siri’s lock-in is getting to be more of a liability than a feature. Siri has been bad for years, and the existence of superior alternatives, and their popularity, is making Siri’s failures into an internet meme… and a bad one.

Foundry

Siri has been bad for years

Siri’s problems started long before the recent boom in LLM-based, generative AI chat companions. Siri has been a disappointment for years.

Here’s a fun little test I like to do from time to time. Open Spotlight on your iPhone (drag down from the home screen) and type in a flight number. United 800, for instance. You’ll get a nicely formatted card of flight info, with an image showing where the plane is in its route, departure, and arrival times, and so on.

Now, ask Siri, “What is the status of United Flight 800.” You’d expect that same card to pop up, right? Apple has had this great flight info card for ages—this is a no-brainer. Instead, you’ll get anything but flight info. You’ll get a general web search, or maybe a news article (as an unreadable notification for the News app), or maybe it will kick the query over to ChatGPT if you have that enabled.

Siri will often deliver wrong historical sports scores, incorrect information on a variety of topics, or perform a different function than the one you asked for. There are dozens of memes, TikTok videos, and social media posts of users asking Siri to do one thing only to have it set a reminder or respond with something else unrelated.

Foundry

It also can’t do things you would expect it to. If I have a timer running for my bread in the oven, and I say, “Siri, add two minutes to the timer” it will start a new 2-minute timer. If I want to do the math myself, looking at how much time is left and adding two minutes, then issuing the command “Siri, change the timer to 8 minutes,” that will work. Frustrating, inconsistent, and unintuitive. I still use Siri every day to add items to my grocery list (which is hit-and-miss and shouldn’t be), set timers, and play music. But this is the stuff that has been solved for a decade. Ten years later I’m doing the same things, only with a glowing border around my iPhone display.

None of this has anything to do with our new age of generative, conversational, LLM-based AI. Like autocorrect, another feature the casual user associated with “AI,” Siri’s general functions seem to be getting worse with age. Maybe they’re not deteriorating and it’s just our expectations that are changing, but either way, Apple has a huge problem on its hands.

With its inability to address obvious Siri problems for the last 5+ years, or to even make basic features like autocorrect work in a way that doesn’t frustrate users, the world’s most valuable company is falling dangerously behind on the biggest technology trend in a generation.

Apple has a couple billion users locked into a tight ecosystem, which buys it a lot of time to right the ship, but it has to demonstrate that it can right the ship. We have to see signs of real, meaningful improvement, and a roadmap that speaks to a company-wide vision of AI that is cohesive, futuristic, and exciting. This is not the kind of thing one can usually expect from a company as famously secretive and incapable of admitting error as Apple.

So I’m left wanting, hoping for a surprise at WWDC or some other event. A clue, hidden in the subtext of a slick marketing video, is that somewhere within Apple Park they have a bright and audacious vision for AI and a plan to deliver it before everyone else does.