Grok Custom Instructions: Where to Find & Set Them in 2026

If you've been using Grok — xAI's flagship AI assistant — for any length of time, you've probably noticed that it doesn't always respond the way you want it to by default. Maybe it's too formal when you need casual output. Maybe it keeps explaining things you already know. Maybe it doesn't know your name, your job, or the kind of content you're creating.

That's exactly the problem custom instructions are designed to solve.

Custom instructions in Grok are persistent preferences you set once, and the AI applies them automatically every single time you start a new conversation. You don't have to repeat yourself. You don't have to re-explain your background. You don't have to paste in a system prompt at the start of every chat. The AI just knows — because you told it, once, in the right place.

In 2026, with AI assistants becoming deeply embedded in daily workflows, custom instructions have gone from being a power-user trick to a fundamental productivity feature. Whether you're a blogger, developer, marketer, student, or creative professional, setting your custom instructions correctly can completely transform how useful Grok is for you on a day-to-day basis.

This guide walks you through exactly where to find the custom instructions setting in Grok's current interface, how to fill it in properly, and how to get the most out of it.

Where Exactly Is the Custom Instructions Setting in Grok (2026)

This is the question most people are searching for, so let's answer it directly before anything else.

To find Grok's custom instructions in 2026, follow these steps:

On Desktop (via grok.com or x.com/grok):

Open Grok in your browser and log in with your X (formerly Twitter) account. Look at the left-hand sidebar — Grok's interface has a collapsible navigation panel. At the very bottom of this sidebar, you'll see your profile icon or account name. Click on it. A small dropdown or settings menu will appear. From there, select Settings. Inside the Settings panel, look for the Personalization section. That's where Custom Instructions lives.

Alternatively, on some versions of the interface, you can click the three-dot menu (⋮) next to your profile, which will also surface the Settings option leading to the same destination.

On Mobile (iOS and Android):

Open the Grok app. Tap the hamburger menu (three horizontal lines) or your profile avatar in the top corner. Scroll down to Settings, then tap Personalization. You'll see the Custom Instructions toggle and text fields there.

On X (Twitter) App with Grok Integration:

If you're accessing Grok through the X app directly, tap your profile picture on the top left to open the sidebar. Scroll down to Settings and Support, then Settings, then look for Grok as a subsection. Inside the Grok settings page, you'll find the Custom Instructions option listed under personalization preferences.

The exact label might vary slightly depending on app version, but the pathway is consistent: Profile → Settings → Personalization → Custom Instructions.

Understanding the Two Fields Inside Custom Instructions

Once you find the Custom Instructions panel, you'll typically see two separate text fields. This two-field structure mirrors what platforms like ChatGPT popularized, and Grok has adopted a similar framework because it works well.

Field One: What Grok Should Know About You

This is your background information field. Think of it as the context Grok needs to give you relevant, personalized responses without you having to explain yourself every time.

What goes here:

• Your profession or area of expertise

• The type of content or tasks you use Grok for most

• Your experience level with the topics you frequently discuss

• Relevant personal context (e.g., "I run a tech blog," "I'm a software engineer with 8 years of experience," "I'm a student studying biochemistry")

• Location if it matters for your queries

• Any ongoing projects or focus areas

The goal of this field is to give Grok a mental model of who you are. The better this is filled in, the more accurately Grok can calibrate its responses to your actual needs rather than assuming a generic user.

Field Two: How Grok Should Respond

This is the behavioral instruction field. This is where you tell Grok how to communicate, not just what to know.

What goes here:

• Preferred tone (formal, casual, direct, friendly, technical)

• Output format preferences (bullet points, prose, numbered lists, tables)

• Length preferences (concise vs. comprehensive)

• Things to always include or always avoid

• Language style (e.g., "avoid jargon," "use plain English," "feel free to use technical terms")

• Whether you want Grok to ask clarifying questions or just attempt an answer

• Persona preferences if any

These two fields together essentially act as a persistent system prompt that shapes every conversation you have with Grok going forward.

How to Enable and Disable Custom Instructions

Inside the Custom Instructions panel, there's a toggle switch at the top. This lets you enable or disable custom instructions globally without deleting what you've written.

This is incredibly useful. If you've set instructions optimized for your work persona but you want to have a casual, unfiltered conversation with Grok, you can flip the toggle off, have your chat, and flip it back on afterward. Your saved instructions remain intact — they're just temporarily paused.

Some users maintain two different sets of instructions by toggling and rewriting, but a cleaner approach is to write flexible instructions that cover most of your use cases rather than switching frequently. We'll cover how to do that in the optimization section below.

Why Grok's Custom Instructions Work Differently Than Other AI Platforms

Grok has some distinct personality traits baked in by xAI — it's designed to be more direct, more willing to engage with edgy or controversial topics, and less prone to over-cautious hedging compared to some of its competitors. This means custom instructions on Grok can be written with a bit more directness without worrying that the model will push back unnecessarily.

Grok also benefits from real-time access to information through its integration with X's data pipeline. Custom instructions don't affect this capability, but they do affect how Grok presents that real-time information to you. If you tell Grok in your instructions that you prefer bullet-point summaries with source context, it will apply that format even to its real-time search results.

Another Grok-specific nuance: the model has different modes in 2026, including a standard mode and a deeper reasoning mode (sometimes called "Think" mode). Custom instructions persist across both modes, so you don't need to re-enter them when switching between Grok's thinking levels.

Step-by-Step: Writing Effective Custom Instructions for Grok

Most people either leave custom instructions blank (leaving massive productivity gains on the table) or fill them with vague, generic text that doesn't actually change how the AI behaves. Here's how to do it properly.

Step 1 — Define Your Primary Use Cases

Before you type anything, think about the top three to five things you actually use Grok for. Write them down. Your custom instructions should be optimized for these use cases, not for every possible scenario.

For example, if you primarily use Grok for writing blog content, coding help, and quick research summaries, your instructions should reflect those three workflows — not try to cover every conceivable thing you might ever ask.

Step 2 — Write Your Background Field Concisely

Aim for three to five sentences that cover who you are, what you do, and what you're focused on. Avoid rambling or over-explaining. The model doesn't need your full biography — it needs enough context to calibrate.

A good example background field:

"I run a tech and gaming blog covering AI, PC hardware, and mobile technology. I have intermediate to advanced knowledge of most topics I ask about, so explanations don't need to start from scratch. I produce long-form SEO articles and also do occasional research for gadget reviews and comparisons. I work in English but sometimes initiate conversations in Bengali."

That's concise, informative, and immediately useful to the model.

Step 3 — Write Your Response Behavior Field With Specificity

Generic instructions like "be helpful and concise" don't do much because Grok already tries to do that by default. The value of this field is in the specific, non-default behaviors you want.

A good example response behavior field:

"Always use clear headings when writing long-form content. Keep explanations direct — I don't need disclaimers or excessive caveats unless they're genuinely important. Prefer structured output with numbered steps for how-to tasks. Don't add filler phrases like 'Certainly!' or 'Great question!' at the start of responses. When writing articles, use H2 headings only, not H3. Avoid inline reference links in article drafts."

These are specific, actionable, and meaningfully different from default behavior.

Step 4 — Test and Iterate

After saving your instructions, have a few conversations that represent your typical use cases. Pay attention to whether Grok is applying your preferences correctly. If something isn't working, go back and adjust the wording. Sometimes the model responds better to slightly differently phrased instructions — for example, "avoid using bold text excessively" might work better than "don't use too much formatting."

Treat your custom instructions as a living document you refine over time rather than a one-and-done setup.

Common Mistakes People Make With Custom Instructions

Even users who find the setting quickly often undercut its effectiveness with some common errors. Here's what to watch out for.

Mistake 1 — Writing Instructions That Contradict Each Other

If your background says "I'm an expert" but your response preferences say "always explain things simply," Grok may get confused about your actual level. Make sure both fields paint a coherent picture.

Mistake 2 — Being Too Vague

"Be helpful and professional" sounds like an instruction, but it's not meaningfully different from Grok's default behavior. Every instruction should describe something specific that you want to be different from the baseline.

Mistake 3 — Overloading the Fields

Custom instructions have character limits, and stuffing them with every possible preference tends to dilute the effect. Prioritize your most important preferences over trying to cover every edge case. Five sharp, specific instructions outperform fifteen vague ones.

Mistake 4 — Forgetting to Toggle Them On

It sounds obvious, but many users write their custom instructions and forget to check that the toggle is actually enabled. Always verify the switch is in the "on" position after saving.

Mistake 5 — Never Updating Them

Your work changes, your projects evolve, your preferences shift. Custom instructions that were perfect six months ago might not reflect your current needs. Make it a habit to revisit and update them every few months.

Advanced Tips: Getting the Most Out of Grok Custom Instructions

Once you've got the basics working, here are some advanced techniques to push your personalization further.

Use Conditional Language for Flexible Instructions

Instead of writing instructions that assume one mode of interaction, you can write conditional preferences that cover multiple scenarios:

"For article writing tasks, use H2 headings only and avoid inline links. For coding tasks, always include comments in the code and explain what each section does. For quick questions, give me a direct answer first, then optional context below."

This kind of instruction gives Grok the flexibility to adjust its behavior based on what you're actually asking for, rather than applying the same rigid format to everything.

Tell Grok What It Should Never Do

Negative instructions are often more powerful than positive ones because they prevent specific behaviors that genuinely annoy you. Think about the things Grok does by default that you consistently find unhelpful, and explicitly call them out:

"Never add unsolicited warnings or disclaimers to creative writing. Never start a response with 'Of course!' or similar affirmations. Never suggest I 'consult a professional' unless the topic genuinely requires it."

Specify Output Templates for Recurring Tasks

If you frequently ask Grok to do the same type of task — write a product description, summarize a news article, draft an email — you can include a template structure in your instructions:

"When I ask you to write a blog article, always structure it with: a compelling intro paragraph, H2 section headings, a FAQ section near the end, and a concluding call-to-action paragraph."

This essentially automates your preferred workflow into every relevant interaction.

Reference Your Expertise Level Precisely

Vague terms like "I'm experienced" or "I know a lot about tech" don't give the model much to work with. Try to be more specific:

"I have five years of hands-on experience with SEO and content marketing. I understand technical concepts like Core Web Vitals, schema markup, and E-E-A-T without needing them explained."

The more precisely you describe your knowledge level in relevant domains, the better Grok can calibrate the depth and vocabulary of its responses.

Grok Custom Instructions vs. Conversation-Level System Prompts

Some power users prefer to paste in detailed instructions at the start of each conversation rather than rely on the persistent custom instructions feature. Both approaches have merit, and they're not mutually exclusive.

Custom instructions are best for stable, long-term preferences that apply across almost all your conversations. They're the baseline.

Conversation-level instructions are best for task-specific requirements that only apply to one session. For example, if you're writing a very specific type of legal document with a narrow set of requirements, you might paste those requirements at the start of that particular conversation rather than bake them into your global settings.

The ideal setup: use custom instructions for your personal baseline (who you are, your general output preferences) and supplement with specific instructions in conversations where you have more precise needs. The two layers stack — Grok will honor both your persistent settings and any additional instructions you provide in the chat itself.

FAQ About Grok Custom Instructions

Does Grok's custom instructions setting apply to the API as well?

No. Custom instructions set through the grok.com or X app interface apply only to your personal use of those consumer-facing products. If you're using Grok through the xAI API, you'd pass a system prompt programmatically in your API calls — the UI-level custom instructions don't carry over.

Can I have different custom instructions for different Grok modes?

As of 2026, custom instructions apply globally across Grok modes (standard and Think mode) rather than being mode-specific. If you want different behaviors per mode, you'd need to either adjust your instructions when switching modes or rely on conversation-level prompts for the difference.

Do custom instructions affect Grok's real-time search results?

They affect how Grok presents and formats information retrieved through real-time search, but they don't change Grok's access to real-time data or which sources it draws from. Your formatting preferences and communication style instructions will apply even when Grok is pulling current information from the web.

Is there a character limit on custom instructions?

Yes. Grok enforces character limits on both fields, though the exact limit can change with platform updates. As a practical rule, aim for concise and specific instructions rather than trying to write a comprehensive document. If you find yourself hitting the limit, that's a sign to prioritize and trim.

Will Grok always follow my custom instructions perfectly?

Not always. Like any language model, Grok sometimes doesn't perfectly honor every preference in every response, especially in complex or multi-part conversations. If you notice a preference being ignored, you can remind Grok mid-conversation, and you may also want to reword the relevant instruction in your settings to make it clearer.

Final Thoughts: Custom Instructions Are the Most Underused Feature in Grok

Most Grok users never touch the custom instructions setting. They keep having the same friction — re-explaining who they are, fixing the same formatting issues, getting responses calibrated for a generic user rather than themselves — and they don't realize there's a single settings page that fixes almost all of it.

Now you know exactly where to find it: Profile → Settings → Personalization → Custom Instructions, whether you're on desktop, mobile, or through the X app.

Take fifteen minutes to fill in both fields thoughtfully. Test a few conversations. Refine the wording until Grok is responding the way you actually want it to. This small investment pays off in every single conversation you have afterward — a faster, more accurate, more personalized AI that doesn't need to be trained from scratch every time you open a new chat.

That's not a minor productivity gain. Over the course of weeks and months of regular use, it's a fundamental upgrade to how you work with AI.