Grok on X: how to use it and how to turn it off

#Quick answer

Grok on X is xAI's AI assistant built into the app. You open it from the Grok icon in the navigation bar on iOS, Android or x.com, and X says it is available in every country where X is available. Free accounts can use it; paying raises the limits. Turning it off is the part most guides get wrong, because there is no single off switch. There are two privacy toggles, a history delete button, and a separate per-language control for the auto-translation feature.

That split matters. The most-linked articles about disabling Grok were written in July 2024 and describe one checkbox. X has reorganised those settings since, and the second toggle is the one people miss.

Last verified 21 September 2026 against X's own help center.

#Where Grok shows up on X

X's help center gives one official entry point: open x.com or the app, tap the Grok icon in the navigation bar, then start typing or pick a suggestion (About Grok, accessed 21 September 2026). That is the full-screen assistant.

In practice Grok reaches you in four more places, and each arrived separately:

• The Grok button next to posts. Added in December 2024 alongside a faster model, it pulls context on the post you are reading, including images, without you typing a prompt (TechCrunch, 13 December 2024).
• Replying with @grok. Since March 2025 you can mention Grok in a reply and the answer posts publicly in the thread (TechCrunch, 7 March 2025). Everyone reading that thread sees it.
• Automatic translation. Rolled out April 2026, this translates posts in other languages for you, powered by Grok (TechCrunch, 8 April 2026).
• The AI image editor, launched at the same time on iOS, which edits photos from a plain-language prompt.

And one place you never see it directly: the ranking system. X's open-sourced For You code is Grok-powered, which we pulled apart in how the X algorithm works. You cannot opt out of that one, because it is the feed itself.

#What Grok can actually do inside X

Split it in two: Grok as a chatbot, and Grok as a reader of X.

As a chatbot it does what you expect: answers, drafting, brainstorming, summarising. Nothing about that is special to X, and grok.com does the same thing in a cleaner window.

The part that only works inside X is real-time access to public posts. X's help page describes Grok deciding, per query, whether to search public X posts and whether to run a live web search. Public posts, post metadata such as engagement and reposts, public Spaces and public profiles are all fair game for that lookup. This is why Grok is genuinely good at "what is everyone saying about this right now" and mediocre at things a general model already handles.

X is direct about the limits too. Its help page warns that Grok "may confidently provide factually incorrect information, missummarize, or miss some context," and advises verifying independently. It also tells users not to put personal, sensitive or confidential information into Grok conversations, about themselves or anyone else. That second warning is worth reading twice before you paste a client email into it.

#Free, Premium, Premium+: what changes

Here is where almost every article gets sloppy. You will find confident claims that free users get ten prompts every two hours, or twenty a day, or thirty per window. None of those numbers appear anywhere on X's own pages.

What X actually publishes is a feature list per tier (About X Premium, accessed 21 September 2026):

So the honest answer to "is Grok free on X" is: yes, at some undisclosed limit that X can change whenever it wants. If you hit that wall regularly, Premium is the cheap fix and Premium+ is the one that changes what Grok can do rather than just how often. Full tier pricing sits in our breakdown of the X Premium price in 2026, and the wider question of what costs money at all is covered in is X free.

One note for anyone weighing Premium+ purely for SuperGrok: X also bundles SuperGrok into Premium Business and Premium Organizations, including for affiliated accounts. If your company already pays for one of those, check before buying a second subscription.

#How to stop Grok training on your posts

Two toggles, same screen, different jobs. X documents both:

1. Open Settings, then Privacy and Safety.
2. Scroll to Data sharing and personalization.
3. Select Grok and Third-party Collaborators.
4. Under Data Sharing, deselect "Allow your public data as well as your interactions, inputs, and results with Grok and xAI to be used for training and fine-tuning."
5. Under Grok Personalization, deselect "Allow X to personalize your experience with Grok."

Step 4 is the one the 2024-era guides cover. Step 5 is a separate control governing whether X shares your profile, top posts, engagement and interests with xAI to tailor Grok to you. Turning off training does not turn off personalization.

Two caveats X states plainly and most summaries drop. First, opting out does not remove Grok or stop Grok-powered features learning through normal use: X gives its own recommendations system as the example. Second, thumbs up and thumbs down feedback is treated as voluntary submission, so a conversation you rate can still be used to train and fine-tune even when you have opted out. If you have opted out on principle, stop rating answers.

If you want the stronger version of this, X says making your account private stops your posts being used for training and stops them being surfaced in response to other people's Grok queries. That is Settings, then Privacy and Safety, then Audience and tagging, then Protect your posts. The trade-offs are real and we walk through them in how to make your X account private.

While you are in that area of settings, it is a reasonable moment to audit which third-party apps still have access to your account. Different mechanism, same category of housekeeping.

#How to delete your Grok conversation history

You can wipe the lot in one action. Settings, then Privacy and Safety, then Data sharing and personalization, then Grok, then Delete Conversation History, then confirm.

X's stated retention after that: deleted conversations are removed from its systems within 30 days, unless it has to keep them for security or legal reasons. There is no selective delete documented, so this is all or nothing. Do the delete after you switch off data sharing, not before, or you leave a gap.

X also answers the obvious follow-up on the same page: no, it does not sell your data.

#How to turn off Grok auto-translation

This one is not in the privacy settings at all, which is why people hunt for it.

Tap the gear icon on a translated post and toggle off automatic translation for that language. It is scoped per language rather than global, so a multilingual timeline may need a few passes before it settles.

Worth knowing before you disable it: auto-translation is also what makes non-English posts searchable and readable in your feed, so switching it off is a preference call, not a privacy one.

#Using Grok to grow on X

A realistic verdict, after using it for research rather than output: Grok is strong at reading X and weak at writing for X.

Reading is the real edge. Asking what a niche is arguing about this week, which posts in a thread carry the actual disagreement, or what an account has been posting lately gives you angles you would otherwise scroll for an hour to find. That is live post access no general-purpose model has.

Writing is where it falls down, and predictably so. Grok writes in the average voice of X, which is the exact voice the feed is saturated with. Posting that output tends to produce replies that read as generated, which is worse than posting less. The engagement math behind that is in our guide to the X algorithm: replies and reposts are weighted far above likes, and generic text earns neither.

If the bottleneck is consistency rather than ideas, that is a scheduling and automation problem rather than an AI-writing one. X-Autopilot runs on your own Mac in your own browser session, which keeps your cookie off a shared cloud server and gives a smaller detection surface than a hosted tool, though no browser automation is risk-free and X's automation rules are the document to read before you set anything running. Automating engagement through a browser is a gray area rather than a sanctioned path, and it is worth being clear-eyed about that.

#Bottom line

Grok is not an add-on you installed. It is stitched into X in five places, and the controls for it are spread across three different screens. If you want it fully dialled back, the sequence is: switch off both toggles under Grok and Third-party Collaborators, delete your conversation history, turn off auto-translation per language on any post that shows it, and consider protecting your posts if training is your actual concern.

If you want more of it, Premium raises limits and Premium+ is the tier that unlocks SuperGrok, Grok Bot and Imagine. And if you are reading this because a 2024 article told you to untick one box, that box is now two.