Tomorrow Next

Apple Is Betting a Smarter Siri Can Erase the Memory of Its AI Trainwreck — But Will Users Care?

Tomorrow Research
Apple Intelligence on Device

Credit: Apple

Apple Is Betting a Smarter Siri Can Erase the Memory of Its AI Trainwreck – But Will Users Care?

The Promise vs. the Reckoning

In June 2024, Apple stood on stage and promised a revolution. Siri would finally understand your screen. It would work across apps. It would remember context. It would feel like talking to an actual intelligent assistant instead of barking commands at a particularly patient toaster.

By January 2025, Apple was pausing notification summaries because its AI had fabricated fake news including a completely invented arrest of Israeli Prime Minister Benjamin Netanyahu and sent those hallucinations to millions of iPhone users.

Now, in March 2026, Apple says it’s ready to try again. The new Siri is coming. It’s different this time. It will understand what’s on your screen. It will integrate with Google’s Gemini models. It will be clever, contextual, and finally worthy of competing with ChatGPT and Google’s Gemini.

The technical case is compelling. The human case ”whether users will actually care enough to upgrade and trust Apple again after 18 months of broken promises is far murkier.

What Went Wrong: The 2024-2025 Disaster

Apple Intelligence was supposed to be Apple’s answer to the AI revolution. In June 2024, the company announced it would ship features that felt genuinely revolutionary: Siri that understood your screen, AI features that ran mostly on your device (privacy preserved), notification summaries that actually made sense.

What could go wrong?

Everything.

The hallucination incident. In January 2025, Apple’s AI notification summaries feature began making things up. BBC News reported that the system fabricated stories: Luigi Mangione, the suspect in a UnitedHealthcare CEO killing, had supposedly shot himself (he didn’t). Three New York Times articles were mangled into a single false notification claiming Netanyahu had been arrested. Users who relied on these summaries to stay informed were instead fed misinformation at the lock screen.

Apple’s response was blunt: disable the feature entirely across iOS, iPadOS, and macOS. It was a rare admission of catastrophic product failure from a company that usually manages its narrative with surgical precision.

The delays. Features promised for iOS 18 (autumn 2024) slipped to “spring 2025,” then to “later in 2025,” and finally to “spring 2026.” That’s an 18-month delay on a feature Apple had publicly committed to delivering. Internally, executives called the situation “ugly” and “embarrassing” (per Bloomberg reporting). Users filed class-action lawsuits.

The trust damage. This is the part that still matters. Apple’s entire value proposition rests on a simple premise: we protect your privacy and we deliver quality. You pay more, but the experience is polished and trustworthy. That premium evaporated in January 2025 when millions of users woke up to fabricated news headlines on their lock screens.

ChatGPT and Google Gemini, meanwhile, had spent 18 months building presence, capability, and user trust. They were already integrated into iPhones via iOS 18.2. Users had moved on.

What the New Siri Actually Does

The 2026 Siri isn’t just a patch. It’s a rewrite from the ground up.

On-screen awareness. This is the cornerstone. Siri can now understand what’s displayed on your iPhone or Mac and act based on that context. Show Siri a purchase confirmation email and ask “find the book Mom recommended,” and it will hunt through your device data to surface the right item. This is genuinely useful-no more explaining what you’re looking at.

Cross-app actions. You can say “increase the volume in Spotify” without opening the app. “Create a reminder from that email.” “Show me the highest-priced item in my cart across all my shopping apps.” These are mundane tasks that become frictionless.

Sustained conversation. The new Siri is powered by large language models (specifically, Google’s Gemini models, in a $1.5 billion/year partnership). That means it can hold conversations, not just execute commands. Ask a follow-up question and Siri remembers context. Ask it to help write an email, and it can draft, refine, and iterate based on your feedback.

Hybrid privacy model. Apple’s strategy here is technically elegant: personal, device-specific requests process locally using Apple’s Neural Engine. Complex queries route to “Apple Private Cloud Compute,” which uses hardware-enforced security guarantees (the servers delete data immediately after processing; third parties can’t intercept). Some tasks might route to Google’s Gemini API. The tradeoff: not all processing is local, but sensitive personal context never leaves your device.

Timeline. iOS 26.4 (rolling out now in March 2026) will include a limited version. The full feature set-especially on-screen awareness-likely arrives in iOS 26.5 (summer 2026) or iOS 27 (fall 2026). Apple committed to shipping something by December 2026. That deadline is hard.

The Competitive Reckoning

This is where Apple’s “comeback” story gets complicated.

ChatGPT in 2026 is not the ChatGPT of 2023. OpenAI has spent two years optimizing voice interaction, web search, image generation, and file analysis. It’s conversational without feeling robotic. It’s integrated into iOS natively (and Android). Millions of users have it as their default first choice for “I need to ask something smart.” Switching costs-especially psychological switching costs-are real.

Google Gemini integrated into Android and Google Workspace is simultaneously more powerful and more convenient for Google ecosystem users. Better multi-step reasoning. Integrated with Gmail, Drive, Docs. If you use a Pixel phone or rely on Google apps, Gemini is already doing most of what the new Siri promises.

Apple’s differentiation is privacy-first architecture and ecosystem depth. On-screen awareness works only because it’s integrated into iOS at the OS level-something ChatGPT can’t easily replicate without deep platform access. The same goes for cross-app actions. These are real competitive advantages.

But they’re invisible to users who are already using ChatGPT.

The honest assessment: Apple is playing catch-up. It will eventually catch up-the team is now led by Amar Subramanya, previously head of engineering for Google Gemini, and Craig Federighi has expanded oversight. The technology is solid. But 18 months of delay means Apple isn’t leading; it’s following with a technically competent alternative.

The Trust Problem Is Structural

Here’s the real story buried beneath the technical improvements: Apple has a trust problem that new features can’t solve.

When a product fails in the way Apple Intelligence failed, users don’t just remember the failure. They adjust their expectations downward and allocate their trust elsewhere. It’s not rational-it’s psychological. You installed ChatGPT on your iPhone for a reason. It works. Why risk switching to something that just admitted it makes things up?

Apple’s executive team understands this. They’ve been signaling that they’re committed to shipping the new Siri on time and delivering actual utility, not just promises. That’s the right messaging. But messaging isn’t enough when you’ve already broken the warranty of trust once.

The new Siri would need to be not just as good as ChatGPT but noticeably better-different enough to justify switching behavior. That’s a higher bar than technical parity. It’s more like 15-20% better on tasks that matter to users. Does it reach that bar? Reviewers will have the final word when it ships, but the betting odds are skeptical.

The real question: Can Apple reclaim AI credibility through actual execution, or is it permanently relegated to “also has an AI assistant” status?

What Reviewers Are Saying

The tech press is cautiously optimistic but skeptical about adoption. The consensus: impressive on paper, wait-and-see on whether users actually care.

MacRumors and Bloomberg report that testing is ongoing and stable, with no major quality regressions. That’s a low bar, we didn’t hallucinate fake news but it’s the bar Apple set for itself.

Gadget Hacks compared ChatGPT voice mode (which has been live for 18 months) to Apple’s upcoming Siri and noted that ChatGPT’s conversational fluency is simply ahead. But they also noted that on-screen awareness is something neither ChatGPT nor Gemini can easily match without OS-level integration.

The underlying sentiment: the technology is respectable. The question is adoption velocity against entrenched competition.

The Bigger Picture: Can Apple Still Lead in AI?

Apple’s historical strength has always been controlling the entire user experience hardware, software, services. It bet that this integration would let it deliver better AI than companies building chat interfaces in isolation.

The 2024-2025 disaster suggests that strategy might be flawed. You can have the best platform integration in the world, but if users don’t trust you, they’ll use someone else’s less-integrated product and accept the friction.

Going forward, Apple’s play is simple: deliver the Siri features on time, make sure they actually work (no hallucinations, no delays), and slowly rebuild trust through reliability. That’s not exciting. It’s not a comeback story. It’s just competent execution.

But competent execution is exactly what Apple needs right now. Execution.

The new Siri might be technically impressive. The real test is whether Apple has learned to ship.

Sources

Exit mobile version