Memory: Teaching Grok to Know You

The Blank Slate Problem

Every time you start a new Grok conversation without memory enabled, Grok knows nothing about you. You are a stranger. If you always want plain language and not technical jargon, you need to say so again. If you work in marketing and not engineering, you need to establish that again. If you prefer bullet points over long paragraphs, you need to ask again.

Grok memory solves this. When memory is on, Grok stores facts about you as they emerge across your conversations and applies them in future sessions — automatically, without you needing to re-introduce yourself each time. It is the difference between a tool that starts from zero every session and one that picks up where you left off.

How Grok Memory Works

Grok stores memories as short notes — individual facts extracted from your conversations. These notes are account-wide and persist across sessions. When you start a new chat, Grok reads your stored memories and uses them as background context before you type a single word.

Memory is not a transcript of your conversation history. Grok does not read back your old chats. It reads the distilled facts that were identified as worth keeping — the preferences and context that make a difference to the quality of future responses.

Memory launched for Grok in April 2025 and is available on both free and SuperGrok plans. Regional availability varies: as of 2026, it is not yet available in parts of the EU and UK due to data privacy regulations. If you do not see memory options in your settings, it may not yet be active in your region.

Finding Your Memory Settings

Memory controls are in Settings → Data Controls → Personalize with Memories. From there you can:

• Toggle memory on or off entirely — Turning it off stops new memories from being stored. Existing memories remain until you delete them.
• View your current memories — See the full list of facts Grok has stored about you.
• Delete individual memories — Remove specific entries without clearing everything.

You can also manage memories directly from the chat interface. When Grok stores something about you, a memory notification may appear in the conversation. Clicking it lets you delete that specific memory without opening settings.

What Gets Remembered

Memory works best for context that repeats across many conversations. The types of facts that improve Grok's responses most consistently:

• Your role or profession — "I am a product manager at a healthcare startup" changes how Grok frames explanations across every future conversation
• Communication style — "I prefer bullet points over paragraphs" or "I want responses under 200 words"
• Domain knowledge level — "I am comfortable with technical depth" or "Explain things as if to a non-technical reader"
• Personal context — Location, dietary preferences, recurring constraints that affect advice across multiple domains

You can also tell Grok explicitly what to remember: Remember that I prefer responses in plain English, not bullet lists. Grok will store this as a memory for future use.

Managing What Grok Knows

Because memories are account-wide, a preference set in one context can carry into unrelated conversations. A few habits that keep your memory profile useful rather than cluttered:

• Review periodically — Open Data Controls occasionally and scan the list. Delete anything that is no longer accurate or that creates cross-context interference.
• Delete after specific tasks — If you tell Grok you are working on a project with unusual constraints, delete that memory when the project ends so it does not colour future responses.
• Be explicit for important preferences — If something matters a lot to how you work, state it clearly ("remember that...") rather than waiting for Grok to infer it.

To clear everything, you can delete individual memories from the Data Controls list. Turning memory off does not automatically delete existing memories — you need to remove those manually if you want a clean reset.

Try This

Open Settings in grok.com, navigate to Data Controls, and check whether memory is available in your region and enabled. If it is on, start a new conversation and tell Grok three things about yourself: your profession or role, how you prefer responses formatted, and one topic where you want a specific level of depth. Check Data Controls after that conversation — you should see those facts listed as stored memories. Start a second, completely unrelated conversation and notice whether Grok's default tone and format already reflect what it remembered. That shift — from zero context to informed context — is the difference memory makes.