How to Create an AI Memory File (Step-by-Step)

Updated January 2026 | 8 min read

You're tired of re-explaining yourself to ChatGPT, Claude, and Gemini every time you open a new chat. You want the AI to remember who you are, what you do, and how you work.

The answer isn't a memory feature. It's a file.

One markdown file. Structured sections. Everything the AI needs to know about you. Save it once, use it forever.

Here's how to build it.

What Is an AI Memory File?

It's a plain text document—usually markdown format—that contains your context. Think of it as a user manual for you.

When an AI reads this file, it knows:

• Who you are (role, industry, business model)
• What you're working on (active projects, clients, deadlines)
• How you communicate (tone, style, banned phrases)
• Your workflows (standard processes, templates, SOPs)
• Domain-specific knowledge (your clients, tools, team structure)

Instead of typing "I'm a real estate agent in Miami, I focus on luxury condos, I prefer direct communication" at the start of every conversation, the AI reads it from the file. Automatically.

Why Markdown?

Markdown is plain text with simple formatting. It's readable by humans and AI. It works everywhere.

You can write it in:

• Notepad (Windows)
• TextEdit (Mac)
• VS Code (cross-platform)
• Obsidian (best for knowledge management)
• Any text editor

No special software required. No proprietary formats. Just text.

Step 1: Create the File

Open your text editor. Create a new file. Name it CLAUDE.md or AI-CONTEXT.md or whatever makes sense to you.

The .md extension means markdown. If your editor doesn't support markdown, use .txt. It'll still work.

Step 2: Start With Identity

First section: who you are. Be specific. Generic descriptions produce generic AI responses.

Bad example:

I'm a consultant who helps businesses with marketing.

Good example:

I'm a fractional CMO for B2B SaaS companies in the $2M-$15M revenue range. I specialize in companies transitioning from founder-led sales to scalable marketing systems. I help them build content engines, optimize paid acquisition, and implement attribution tracking. Average engagement: 6-12 months. Average retainer: $8K-$15K/month.

The second version gives the AI context it can use. It knows your target market, your service model, your pricing, your engagement length. That affects how it drafts proposals, writes emails, suggests strategies.

Example structure:

## Identity

I'm [name], [role] at [company/industry]. I [what you do specifically].

My focus: [target market/niche]
My business model: [how you make money]
Experience: [years, relevant background]

Step 3: Define Your Voice

How should the AI write for you? This section controls tone, structure, and style.

Example:

## Voice

**Tone:** Direct, conversational, no corporate jargon. Use contractions. Sound like a person, not a press release.

**Structure:** Lead with the main point. Then supporting details. No long intros.

**Format:**
- Bullet lists over paragraphs when listing items
- Short sentences (under 25 words when possible)
- Active voice, not passive
- Specific numbers, not vague claims

**Banned words/phrases:**
- Solutions, leverage, synergy, ecosystem
- "Moving forward," "circle back," "touch base"
- "Cutting-edge," "world-class," "innovative"

**Writing style:**
- Use "you" and "I," not "one" or "we" (unless referring to a team)
- End sentences when the idea is complete, don't keep going
- If it sounds like marketing copy, rewrite it

This gives the AI concrete rules. It won't produce generic corporate fluff because you've explicitly banned it.

Step 4: List Your Projects

What are you working on right now? Active clients, internal initiatives, launches, campaigns.

Example:

## Active Projects

**Client: Acme SaaS (CRM platform for real estate)**
- Engagement start: Jan 2026
- Current phase: Content strategy + SEO audit
- Deliverable due: Feb 28 (topical authority map + Q1 content calendar)
- Primary contact: Jennifer Lee (VP Marketing)

**Client: Pacific Logistics (supply chain software)**
- Engagement start: Nov 2025
- Current phase: Paid acquisition optimization
- Focus: Google Ads + LinkedIn, attribution setup
- Primary contact: Marcus Chen (Founder/CEO)

**Internal: Course launch "AI for Marketers"**
- Launch date: March 15, 2026
- Status: Module 3 of 6 complete
- Platform: Teachable
- Pricing: $497

When you ask the AI to draft a status email for Acme, it knows the project phase, the deadline, the contact. You don't re-explain it.

Step 5: Add Domain Context

This is where you document the specific knowledge the AI needs about your work. Clients, tools, workflows, team structure.

Example for a consultant:

## Clients

**Acme SaaS**
- Product: CRM for real estate agents
- Stage: Series A, $5M ARR, 40 employees
- My role: Content + SEO strategy
- Contact: Jennifer Lee (VP Marketing), jlee@acme.com

**Pacific Logistics**
- Product: Supply chain visibility platform
- Stage: Seed, $800K ARR, 12 employees
- My role: Paid acquisition + attribution
- Contact: Marcus Chen (CEO), marcus@pacificlogistics.io

## Tools I Use

- Project management: Notion
- Communication: Slack (clients), Email (prospects)
- Analytics: Google Analytics 4, Looker Studio
- SEO: Ahrefs, Screaming Frog
- Paid ads: Google Ads, LinkedIn Campaign Manager

## Standard Workflows

**New client onboarding:**
1. Kickoff call (45 min): goals, timeline, access/credentials
2. Audit phase (week 1-2): current state analysis
3. Strategy doc (week 3): recommendations + roadmap
4. Execution begins (week 4+)

**Monthly client reporting:**
- Sent first Monday of each month
- Sections: metrics, wins, challenges, next month plan
- Format: Google Doc, then PDF export
- Always include 3-month trend graphs

This level of detail means the AI can draft a client report without you specifying the format, the sections, the timing. It's all documented.

Step 6: Document Your Preferences

Small details that shape how the AI helps you.

Example:

## Preferences

**Dates:** Use YYYY-MM-DD format (2026-01-28, not Jan 28, 2026)

**Names:** Always include full context on first mention. "John Smith (Acme SaaS, VP Product)" not just "John."

**Email style:**
- Subject lines: under 50 characters, no clickbait
- Greeting: "Hi [Name]," (not "Hello" or "Hey")
- Closing: "Thanks," or "Best," (not "Sincerely" or "Regards")
- Always include clear next step or call to action

**Meeting notes format:**
- Date + attendees at top
- Decisions made (bullet list)
- Action items (who, what, when)
- Open questions
- Next meeting date

These are the tiny things you'd normally have to correct after the AI drafts something. Document them once, never correct them again.

Step 7: Include Templates

If you send similar emails or documents repeatedly, include templates.

Example:

## Email Templates

### New prospect outreach

Subject: [Specific problem] — quick question

Hi [Name],

I saw [specific detail about their company/content].

[One sentence connecting that detail to a problem you solve.]

I work with [type of companies] to [outcome you deliver]. [One sentence case study or result].

Worth a 15-minute call to see if there's a fit?

[Calendar link]

Thanks,
Victor

---

### Client status update

Subject: [Client name] update — [Month YYYY]

Hi [Name],

Quick update on what we shipped this month:

**Completed:**
- [Item 1]
- [Item 2]

**In progress:**
- [Item 3]
- [Item 4]

**Next month:**
- [Item 5]

Any questions, let me know. Otherwise, we'll keep moving.

Thanks,
Victor

When you ask the AI to draft an outreach email or status update, it uses your template. You're not starting from scratch.

Step 8: Save and Store It

Save the file somewhere you can access easily. Recommendations:

Option 1: Project directory
If you use Claude Code or another AI that reads local files, save CLAUDE.md in your main working directory. It'll read it automatically.

Option 2: Obsidian vault
If you're building a larger knowledge system, create an Obsidian vault and save your context file there. You can link to other notes, build a multi-file system.

Option 3: Cloud storage
Save it in Google Drive, Dropbox, or OneDrive so you can access it from any device. Just make sure you can download it when needed.

Step 9: Use It

How you use your context file depends on the AI platform:

Claude Code: Save CLAUDE.md in your working directory. Claude Code reads it automatically every session. No manual steps.

ChatGPT (Advanced Data Analysis): Upload your context file at the start of each session. It'll reference it for the entire conversation.

Claude Projects: Upload your file to a project. It persists across all chats in that project.

Any AI (manual): Copy the contents of your file and paste it at the start of important conversations. Tedious but functional.

Step 10: Update It Regularly

Your context file isn't static. As your work evolves, update it:

• New client? Add them to the Clients section.
• Project wraps up? Remove it from Active Projects.
• Preferences change? Edit the Voice section.
• New workflow? Document it.

Set a reminder to review your context file monthly. Takes 10 minutes. Keeps the AI accurate.

Example: Complete Starter File

Here's what a minimal context file looks like:

# AI Context - Victor Romo

## Identity

I'm Victor Romo, founder of Scale With Search. I teach SEO and AI workflows to service businesses and consultants. My clients are typically solo practitioners or small teams (under 10 people) who want to rank for high-intent keywords and use AI to scale content production without sacrificing quality.

## Voice

- Tone: Direct, conversational, no corporate jargon
- Structure: Lead with the main point
- Format: Bullet lists, short sentences, active voice
- Banned words: solutions, leverage, ecosystem, cutting-edge

## Active Projects

- Client onboarding system redesign (due Feb 2026)
- SEO course Module 3 (launching March 2026)
- Weekly newsletter (published Thursdays)

## Preferences

- Dates: YYYY-MM-DD format
- Email greeting: "Hi [Name],"
- Always include next steps in emails

That's 150 words. Enough to make every AI conversation better.

From there, expand. Add clients. Document workflows. Build the system that fits your work.

Common Mistakes

Being too vague. "I'm a consultant" doesn't help. "I'm a fractional CFO for SaaS companies preparing for Series A fundraising" does.

Including outdated information. If your context file says you're working on a project that wrapped up three months ago, the AI will reference it. Update regularly.

Overcomplicating it. Start simple. You don't need 50 pages on day one. Start with 300 words. Expand as you need it.

Not using it. The file only helps if you actually use it. If you're still opening ChatGPT without loading your context, you're not getting the benefit.

Start Today

Open a text editor. Create CLAUDE.md. Write 300 words:

• Who you are
• What you do
• How you want the AI to communicate

Save it. Use it in your next AI conversation. See the difference.

Then expand. Add projects. Document workflows. Build the knowledge base that makes AI actually useful.

One file. One afternoon. Persistent memory that doesn't forget.