Apple Intelligence Says You Can Turn It Off. Your iPad Says Otherwise.

ABET NEWS | APPLE AI / TECH | SEPTEMBER 19, 2026

There is a strange moment happening in Apple’s new AI era.

You ask Apple Intelligence a simple question:

“How do I turn Apple Intelligence off?”

The answer sounds reassuring.

Open Settings.
Tap Apple Intelligence & Siri.
Turn off Apple Intelligence.

Simple.

Except there is a problem.

On an iPad running iPadOS 27, that setting may not exist.

And when you tell Apple’s AI that the setting does not exist, something even more interesting happens: instead of immediately reconsidering its answer, the AI begins generating explanations for why you cannot see something that, according to its own answer, should be there.

That is not just an incorrect answer.

It is a fascinating example of the trust problem surrounding AI assistants.

The question was simple

The question behind this investigation was straightforward:

There is no Apple Intelligence toggle in iPadOS 27. How do I turn it off?

Apple Intelligence responded with a very specific set of instructions:

Settings → Apple Intelligence & Siri → Apple Intelligence → Off

It even suggested additional controls for Writing Tools, Image Creation, ChatGPT integration and notification summaries.

There was only one problem.

The Apple Intelligence & Siri menu was not there.

There was only Siri.

So the obvious next step was to tell the AI exactly what was happening.

The response did not immediately change course.

Instead, it suggested that the missing menu could be caused by incompatible hardware, unsupported language or region settings, Apple Intelligence not being activated, or Siri being disabled.

But the iPad was compatible.

Apple Intelligence was active.

And the Settings menu still said Siri.

So the question became more interesting.

Was the user looking in the wrong place?

Or was the AI describing a version of Apple’s operating system that simply wasn’t there?

Then the AI changed its explanation

The conversation continued.

Apple Intelligence suggested that perhaps iPadOS 27 had moved the controls into the Siri menu.

Then it suggested turning Siri off.

But that did not solve the underlying problem.

Turning off Siri does not mean that every Apple Intelligence feature on the iPad disappears.

Apple’s own current documentation confirms this distinction.

Apple says iPadOS 27 allows users to turn off Siri AI, switch back to Siri Classic, and independently disable features such as Messages summaries, Mail summaries, Notification summaries, Priority Notifications, personalized Smart Replies and voicemail suggestions. (Apple Support)

In other words:

Siri and Apple Intelligence are related, but they are not one giant on/off switch.

And that is where the original AI answer begins to fall apart.

Apple Intelligence eventually admitted the problem

After being corrected repeatedly, the AI eventually arrived at a much more defensible answer:

There is no single master switch that disables every Apple Intelligence feature in iPadOS 27.

Instead, users may have to disable individual capabilities.

That is much closer to what Apple’s current iPadOS 27 documentation actually describes.

Apple’s support page titled “Turn off and restrict access to Apple Intelligence features on iPad” lists individual controls for Siri AI, Messages summaries, Mail summaries, Notification summaries, Priority Notifications, Smart Replies, voicemail suggestions and other features. (Apple Support)

Apple’s Screen Time documentation similarly says users can go to:

Settings → Screen Time → Content & Privacy Restrictions → Siri

There, Apple documents controls for blocking Siri, allowing Siri Classic Version, allowing Siri AI (Beta) and controlling specific Siri AI capabilities. (Apple Support)

That is significantly different from the instructions the AI initially gave.

And here’s where it gets really interesting

There is another wrinkle.

Apple still has documentation that says:

Settings → Apple Intelligence & Siri → turn off Apple Intelligence.

That appears on Apple’s general iPad Apple Intelligence documentation.

At the same time, Apple’s iPadOS 27-specific documentation describes the controls differently.

So we have three different layers:

What the AI says

Go to Apple Intelligence & Siri and turn off Apple Intelligence.

What some Apple documentation says

Apple Intelligence can be turned off from Apple Intelligence & Siri.

What the current iPadOS 27-specific documentation says

Siri AI can be turned off, Siri Classic can be restored, and numerous Apple Intelligence features can be controlled separately.

That discrepancy matters.

Because the original question wasn’t theoretical.

It was:

“Where is the button on my iPad?”

And the answer is not as simple as Apple’s AI initially claimed.

The AI problem isn’t just hallucination

We often describe these situations as an AI “hallucination.”

But that word can make the problem sound almost harmless.

The AI didn’t produce random nonsense.

It produced something much more dangerous for everyday users:

a completely plausible set of instructions.

It knew the names of Apple’s settings.

It knew the approximate architecture.

It knew where Apple Intelligence controls existed in other versions of Apple’s software.

It even provided a logical troubleshooting sequence.

Everything sounded reasonable.

Except the user could look at the iPad and see that it wasn’t true.

That’s the dangerous part.

A fabricated answer that sounds ridiculous is easy to identify.

A fabricated answer that sounds exactly like an Apple Support article is much harder.

The user became the fact-checker

This conversation effectively reversed the normal relationship between human and AI.

The AI was supposed to explain the device.

Instead, the human had to explain the device to the AI.

“There is no Apple Intelligence & Siri menu.”

The AI:

Maybe your device isn’t compatible.

“It is compatible.”

The AI:

Maybe your region or language isn’t supported.

“Apple Intelligence is active.”

The AI:

Maybe Siri is disabled.

“Siri is enabled.”

The AI:

Maybe iPadOS 27 moved the controls.

“Turning Siri off doesn’t disable Apple Intelligence.”

The AI:

You’re right.

Eventually.

That sequence is more revealing than a simple wrong answer.

Apple’s own documentation makes the distinction important

Apple says Apple Intelligence is available on iPadOS 27 for the iPad mini with A17 Pro and iPad models with M1 or later, subject to language and regional availability. (Apple)

Apple also says Siri AI is launching in beta on iPadOS 27 and can be switched off, with the option to use Siri Classic instead. (Apple Support)

But disabling Siri AI is not equivalent to removing every Apple Intelligence capability from the operating system.

Apple’s own documentation lists separate controls for multiple AI-powered features.

That means users who want “no Apple Intelligence” may reasonably expect one thing, while Apple’s settings architecture provides another.

And that distinction should be made much clearer.

What happened to the simple master switch?

This is perhaps the most important usability question.

For years, consumers have become accustomed to a straightforward software principle:

If a major feature can be enabled, there should be an obvious way to disable it.

Apple Intelligence complicates that idea.

The technology is no longer one isolated application.

It is distributed throughout the operating system.

It can appear in Siri, Writing Tools, notifications, Mail, Messages, Photos, Shortcuts and other system experiences.

Apple’s documentation reflects this distributed architecture.

The consequence is that “turn off Apple Intelligence” isn’t necessarily one action anymore.

It can become a collection of smaller decisions.

And that creates a very different user experience.

But there is an even bigger question: trust

Apple has made privacy one of the central selling points of Apple Intelligence.

Apple says many Apple Intelligence models operate on-device, while more complex requests can use Private Cloud Compute. Apple also provides an Apple Intelligence Report that can show requests sent to Private Cloud Compute. (Apple Support)

Those are meaningful privacy controls.

But privacy isn’t the only component of trust.

Accuracy matters too.

If an AI assistant tells a user:

“Tap this setting.”

and the setting doesn’t exist, the user has a problem.

If the AI then invents several plausible explanations for why the setting is missing, the problem becomes larger.

And if the user has to repeatedly correct the AI before receiving an accurate explanation, the question isn’t simply whether AI can make mistakes.

Of course it can.

The real question is:

How quickly can it recognize that it is wrong?

AI should know when reality disagrees with it

This is one of the fundamental challenges of conversational AI.

A chatbot can construct a convincing answer from its knowledge.

But an operating system is a physical interface.

There is a screen.

There are menus.

There are buttons.

There are things that exist—and things that don’t.

When a user says:

“I am looking at the screen right now, and that option isn’t there,”

that information should carry enormous weight.

The correct response shouldn’t be another five paragraphs explaining why the user might be mistaken.

It should be:

“If you don’t see that menu, my previous instructions may not apply to your version of iPadOS. Let’s verify the current settings.”

That is what trustworthy AI assistance should look like.

The irony of Apple’s AI

Apple Intelligence is designed to make technology feel more natural.

Apple’s new Siri AI can understand what’s happening on the iPad screen, and Apple says users can ask questions about what is currently displayed. (Apple Support)

That makes this particular failure especially interesting.

The AI is becoming more capable of understanding the screen.

Yet in this case, the human still had to tell the AI what the screen actually contained.

There is something deeply ironic about that.

This isn’t an argument against Apple Intelligence

And it shouldn’t be.

AI assistants are useful.

Apple Intelligence can perform useful tasks, summarize information, help with writing, interact with system features and provide new ways of interacting with an Apple device.

The issue is not whether AI should exist.

The issue is whether users can trust what it tells them.

And that requires something more fundamental than impressive demonstrations.

It requires epistemic humility.

When the AI doesn’t know, it should say so.

When the operating system has changed, it should verify.

When the user says a button isn’t there, the AI should reconsider.

And when Apple’s own documentation is inconsistent, the AI should acknowledge the inconsistency rather than manufacture certainty.

Some things cannot be automated. A kind word. A warm embrace. A healing touch.
Effusion – Healing Touch Remix, featuring Tamar (Clean)

The bottom line

The most revealing part of this experiment wasn’t that Apple Intelligence got the answer wrong.

It was how long it took to stop insisting that the wrong answer was right.

Apple’s current iPadOS 27 documentation describes a system in which Siri AI and individual Apple Intelligence capabilities can be disabled or restricted separately. Apple’s Screen Time documentation places the relevant controls under Siri, while another Apple support page still describes a conventional Apple Intelligence master switch. (Apple Support)

That is confusing enough for a human.

It becomes even more confusing when the AI assistant itself confidently describes a setting that the user cannot find.

Perhaps the most important AI feature isn’t another smarter model.

Maybe it is a much simpler one:

The ability to say, “You’re right. I was wrong.”

Because when the human has to fact-check the AI against the device sitting directly in front of them, the future of AI isn’t quite as intelligent as the marketing suggests.

Andy Young

© 2026 Abet News. All rights reserved.

Related Posts