How to Turn Off AI in Gmail

Your inbox started finishing your sentences, summarizing your threads, and offering a chatbot you never asked for. If you want to know how to turn off AI in Gmail, the short answer is: you can disable most of it, but the process is fragmented across multiple settings pages, it takes some useful non-AI features down with it, and parts of the AI interface will persist anyway. Here is every switch, what each one actually controls, and what "off" does not cover.

Key Takeaways

• There is no single "disable all AI" toggle. You need to uncheck at least two separate settings, and on enterprise accounts, your admin may need to act.
• Disabling "Smart features and personalization" kills AI summaries, Smart Compose, Smart Reply, and nudges, but also removes non-AI functions like spellcheck.
• The AI side panel on personal accounts is tied to a paid subscription; collapsing it or canceling the subscription are your only options, as there is no dedicated off-switch.
• A January 2026 redesign is embedding AI directly into the inbox UI rather than keeping it in a separate panel, which will make visual opt-out harder over time.
• EU, UK, and Japanese users were opted out by default; US users were opted in. Check your settings regardless.

Why does Gmail have AI features enabled by default?

Because the provider switched from opt-in to opt-out. In late 2025, AI access was rolled out to most US consumer and Workspace accounts by default, with the relevant toggles buried three menu layers deep. Users in the EU, UK, and Japan were opted out by default, which tells you something about how different regulatory environments produce different product defaults.

The provider's official account on X has stated that email content is not used to train its AI model. Critics have pointed out that this is a narrow denial: it does not address whether AI processes or analyzes your email content for other purposes, like generating summaries or suggested replies. The distinction between "training on" and "processing" matters if you care about what happens to the text of your emails at inference time.

What does the "Smart features and personalization" toggle actually control?

It controls the largest bundle of AI functions in a single checkbox. Unchecking it disables Smart Compose, Smart Reply, nudges, automatic email categorization, and in-thread AI summaries. It is the closest thing to a master switch, and it is the first thing you should turn off.

The catch: this toggle also disables features that predate the current AI wave. Spellcheck goes away when you uncheck it. So does automatic categorization into Primary, Social, Promotions tabs. These are features most people consider basic email functionality, not AI. Bundling them together means the provider has created a cost to opting out. You lose spellcheck to stop getting AI summaries. That is a design choice, not a technical constraint.

How do I find this setting?

Open Gmail on desktop. Click the gear icon, then "See all settings." Navigate to the "General" tab, scroll down to "Smart features and personalization," and uncheck the box. Then scroll a bit further to "Smart features and personalization in other products" and uncheck that too. Click "Save Changes" at the bottom. You need to do this per account; there is no global toggle that covers all your accounts at once.

The second checkbox ("in other products") controls whether your email data can be used by the provider's other services to personalize things like maps, travel suggestions, or the assistant. If you only uncheck the first one, your email data may still flow into other products.

How do I remove the AI side panel?

You mostly cannot. On personal accounts, the AI side panel (the star or sparkle icon in the upper right) only appears if you subscribe to the provider's paid AI tier. Your options are: collapse it by clicking the icon so it tucks away, or cancel that subscription entirely. There is no setting within Gmail that says "hide side panel."

For Workspace accounts, the side panel may persist regardless of what you do individually. Your organization's admin controls AI access across the domain, and individual users often cannot fully disable it without admin intervention. If you are on a managed Workspace account and the AI panel bothers you, your next step is an email to IT, not a settings page.

What about the inline AI features replacing the side panel?

They are harder to turn off because they are not a separate UI element. Starting in January 2026, the standalone AI side panel began being phased out for personal subscribers in favor of AI woven directly into the inbox: context-aware suggested replies, "Help Me Write" prompts inside the compose window, enhanced proofreading. These inline features are not a collapsible panel you can tuck away. They live inside the core email interface.

This is a structural shift worth paying attention to. When AI is a bolt-on panel, you can at least visually ignore it. When AI is embedded into compose, reply, and thread views, "turning it off" means disabling the Smart features toggle (with all the collateral damage described above) or just learning to ignore the suggestions. The provider is making AI inseparable from the product rather than offering it as a module. The direction is clear: future opt-outs will get harder, not easier.

Does Gmail's own settings page fully remove all AI elements from the interface?

No. Even after disabling Smart features, the AI button often remains visible in the top navigation bar. It may not do anything functional once you have toggled the settings off, but it sits there. For some users this is a minor annoyance. For others it is a reminder that "off" in this context means "mostly off, with a persistent visual nudge to turn it back on."

Third-party browser extensions have emerged specifically to hide these leftover UI elements. Tools like CMDK exist to remove the AI button and other interface artifacts that the provider's own settings do not address. The fact that such tools have a user base tells you something about unmet demand for a genuinely AI-free inbox.

Step-by-step: the full opt-out sequence

Here is the complete process, combining everything above into one checklist. This covers both consumer and Workspace accounts where possible.

1. Disable Smart features and personalization. Gmail settings → General tab → Uncheck "Smart features and personalization." This kills Smart Compose, Smart Reply, nudges, automatic categorization, and AI summaries. It also kills spellcheck.
2. Disable Smart features in other products. Same settings page, just below the first toggle. Uncheck "Smart features and personalization in other [provider] products." This stops your email data from flowing into maps, travel, and the assistant.
3. Save changes. Scroll to the bottom. Click "Save Changes." Gmail will reload.
4. Collapse or remove the AI side panel. If you see a star or sparkle icon in the upper right, click it to collapse the panel. If you are on a paid AI subscription and want it gone entirely, cancel that subscription.
5. Repeat per account. These settings are per-account. If you have a personal account and a Workspace account, you need to do this twice.
6. For Workspace users: contact your admin. If AI features persist after you have changed your individual settings, your Workspace admin may need to disable AI access at the organizational level.
7. Optional: install a browser extension. If the AI button remains in the top bar and it bothers you, a tool like CMDK can hide it.

Total time: about three minutes per account, assuming you know where to look. Which you now do.

What are the side effects of disabling Smart features?

You lose more than AI. Here is what goes away when you uncheck that first toggle:

• Spellcheck. The built-in spellcheck is bundled under Smart features. Disabling one disables the other.
• Automatic categorization. Your Primary, Social, Promotions, and Updates tabs stop being automatically sorted. Everything lands in one undifferentiated inbox.
• Smart Compose and Smart Reply. The gray predictive text while typing and the three suggested reply buttons at the bottom of emails both disappear.
• Nudges. Those "Received 3 days ago. Reply?" prompts at the top of your inbox go away.
• AI-generated summaries. Thread summaries and the side panel's ability to answer questions about your inbox stop working.

Whether this tradeoff is acceptable depends on how you use email. If you rely on tab-based sorting or spellcheck, losing them is a real cost. If you mostly use a desktop client with its own spellcheck and prefer a single-stream inbox anyway, you may not miss anything.

Is there a legal basis for this being opt-out instead of opt-in?

That question is being tested in court, inconclusively. A class action (Thele v. the provider) was filed alleging that the silent switch from opt-in to opt-out AI processing violated user privacy. On July 7, 2026, a federal judge dismissed the case, ruling that the plaintiffs had not described a concrete enough injury to establish standing. They were given 21 days to amend their complaint.

The dismissal did not say the practice was fine. It said the plaintiffs had not yet shown they were harmed in a way the court could act on. This is a pattern in privacy litigation: the burden falls on users to prove harm after the fact, rather than on providers to obtain consent before processing. For users, the practical takeaway is that courts are not currently stopping default-on AI processing, so the settings page is your only real recourse.

Why does "off" not really mean off?

Because the settings control functionality, not data flow. When you uncheck Smart features, the visible AI behaviors stop: no more suggested replies, no more summaries. But the provider's narrow public statement about not using email content to train AI does not address what other processing happens on the backend. The setting removes the user-facing output. It does not give you a guarantee about what happens to your email text on the provider's servers.

This is a fundamental limitation of toggling features off inside a platform you do not control. The toggle governs what the product shows you. It does not govern the product's architecture. If your concern is not just "I find suggestions annoying" but "I do not want my email content processed by AI infrastructure at all," a settings toggle within that same infrastructure cannot fully address that concern. You would need to use a different email provider, or at minimum, a client-side encryption layer.

What about mobile?

The Smart features toggles apply across platforms, so disabling them on desktop should propagate to the mobile app. However, the AI side panel has a different rollout timeline on Android than on web, and some inline AI features may appear on mobile even when the side panel has been phased out on desktop. If you have changed settings on desktop and still see AI suggestions on your phone, open the Gmail app, go to Settings → your account → General settings, and verify that Smart features toggles are unchecked there as well.

What does this tell you about consent architecture?

The design of this opt-out is worth studying on its own terms. Consider the choices:

AI features were enabled by default for US users, while EU users were opted out. The opt-out requires navigating multiple settings pages. Disabling AI also disables non-AI features (spellcheck) that most users consider essential. Even after toggling off, visual AI elements persist in the UI. And the provider is actively moving from a collapsible side panel to inline AI that cannot be visually separated from the core product.

Each of these is a design decision that increases friction for opting out. Taken individually, each one has a plausible product rationale. Taken together, they form a consent architecture that structurally favors keeping AI on. The default is on. The off switch has collateral costs. The visual reminder stays. And the future direction makes the AI harder to distinguish from the product itself.

This is not unique to email. It is a pattern across products that monetize user data or engagement: make the default the thing you want users to do, make the alternative costly, and gradually integrate the feature so deeply that separating it becomes impractical. If you build products, this is worth understanding. If you use products, it is worth noticing.

Are there alternatives to toggling settings inside the same platform?

Yes. If your goal is an inbox that does not process your email content with AI at all, the most direct solution is to use an email provider that does not offer AI features (or offers them strictly as opt-in). Several privacy-focused providers exist. The tradeoff is leaving the provider's ecosystem, which has real costs if you depend on its calendar, drive, or other integrated services.

If you want AI when you choose it but want it separated from your email entirely, that is a different problem. You can use a standalone AI assistant that has no access to your inbox unless you explicitly paste content into it. Selina works this way: it remembers context across conversations but only knows what you give it, and your memory is encrypted at rest. It is not reading your inbox in the background. You bring content to it; it does not pull content from you.

The underlying question is whether AI should be woven into the infrastructure you depend on (where opting out means losing functionality) or offered as a separate tool you invoke when you want it (where not using it costs you nothing). These are different architectures with different implications for user control. The current trajectory of major email platforms is clearly toward the former.

A quick note on Workspace admin controls

If you manage a Workspace domain and want to disable AI for your organization: the admin console has controls for enabling or disabling AI features across your domain. Individual users on Workspace accounts often cannot fully disable AI without admin action, so if you are getting requests from your team, this is your responsibility, not theirs. The admin console path changes periodically, so check the provider's current Workspace admin documentation rather than relying on a screenshot from six months ago.

Summary of what actually works

Disabling Smart features and personalization (both toggles) removes the majority of AI behaviors from your inbox. It also removes spellcheck and tab sorting. The AI side panel can be collapsed but not fully removed via settings; canceling a paid AI subscription removes it on personal accounts. Residual AI buttons in the navigation bar require a browser extension to hide. Workspace users may need their admin to act. And the ongoing shift to inline AI means the surface area for opt-out is shrinking over time.

None of this is complicated. It is just fragmented, which is a different kind of problem. The information exists. The toggles exist. The friction is the point.

If you want an AI assistant that only knows what you tell it, and where delete means gone: start a free 7-day trial, no card required.

Michael C.

Founder & Principal Engineer, Selina Labs

Michael builds Selina, a privacy-first AI that remembers you across conversations. He ships security-sensitive AI in production — real attacks, real fixes, measured in minutes and dollars — and writes about privacy, security, and LLMs from that seat. Top Rated Plus and expert-verified on Upwork.

• www.upwork.com
• x.com
• medium.com
• selinalabs.io