Adding Knowledge Files

Knowledge files give your GPT access to specific information it wouldn't otherwise know. This is what transforms a general assistant into a specialist that knows YOUR content.

What Are Knowledge Files?

Knowledge files are documents you upload to your GPT. When users ask questions, your GPT can search these files for relevant information and use it in responses.

Without knowledge files: Your GPT only knows what ChatGPT was trained on With knowledge files: Your GPT can reference your specific documents, guides, and data

When to Use Knowledge Files

Knowledge files are ideal for:

Use Case	Example File
Company information	Employee handbook, product docs
Reference materials	Style guides, templates, checklists
Specialized knowledge	Research papers, technical specs
Structured data	FAQs, pricing tables, schedules
Your own content	Blog posts, course materials

Supported File Types

Custom GPTs accept these formats:

• Documents: PDF, DOCX, TXT, MD
• Spreadsheets: CSV, XLSX
• Presentations: PPTX
• Images: PNG, JPG, GIF, WEBP
• Code: PY, JS, TS, HTML, CSS, JSON

Limits: Up to 20 files, 512 MB total per GPT.

How to Upload Files

1. In the GPT Builder, go to the Configure tab
2. Scroll to the Knowledge section
3. Click Upload files
4. Select your files
5. Wait for processing to complete

Preparing Effective Knowledge Files

Not all files work equally well. Follow these tips:

Use Clear, Searchable Text

Good: Plain text, well-formatted documents Bad: Scanned images of text, handwritten notes

Organize with Headers

Headers help your GPT find relevant sections quickly.

Be Specific and Complete

Include details your GPT will need:

Vague	Specific
"Flexible pricing"	"$49/month, $449/year, 15% team discount"
"Contact support"	"support@company.com, Mon-Fri 9am-5pm EST"
"Various options"	"Option A: ..., Option B: ..., Option C: ..."

Use FAQ Format for Common Questions

Referencing Files in Instructions

Tell your GPT how to use uploaded files:

What NOT to Upload

Avoid uploading:

• Sensitive data: Passwords, API keys, personal information
• Copyrighted content: Books, paid courses (unless you own rights)
• Outdated information: Old price lists, deprecated docs
• Huge files: Break large documents into focused sections

Testing Knowledge Retrieval

After uploading, test that your GPT can find and use the information:

1. Ask questions that require file knowledge
2. Ask for specific details from the documents
3. Try variations of how a user might phrase questions
4. Check that responses accurately reflect your files

Example test questions:

• "What's the refund policy?" (tests FAQ retrieval)
• "How much does the pro plan cost?" (tests pricing data)
• "What are the steps to reset my password?" (tests procedural content)

Updating Knowledge Files

Knowledge can become outdated. Plan to:

• Review files quarterly
• Update pricing and dates
• Remove deprecated information
• Add new FAQs based on user questions

To update: Delete the old file, upload the new version.

Example: Building a Customer Support GPT

Files to upload:

1. product-faq.md - Common questions and answers
2. pricing-plans.csv - Current pricing details
3. troubleshooting-guide.pdf - Technical solutions
4. company-policies.txt - Return, refund, and support policies

Instructions addition:

"For all product questions, first check your knowledge files. When answering about pricing, use exact figures from pricing-plans.csv. For technical issues, follow the steps in troubleshooting-guide.pdf."

Key Takeaway

Knowledge files transform a generic GPT into a specialized expert on your content. Upload clear, well-organized documents, tell your GPT how to use them, and test that information retrieval works. Next, we'll create conversation starters to help users get started quickly.