• Mission statement and company values
• Brand guidelines and style guide 
• Detailed product/service descriptions/menu
• Technical specifications
• Pricing information
• Demographics and Target Market
• Frequently Asked Questions 
• Customer reviews and testimonials
• Common pain points, objections, and motivations 
• Past successful campaigns
• Email marketing templates
• Ad copy examples
• Social media posts
• Press releases 
• Sales scripts
• Sales history and goals
• Common objections and responses
• Case studies 
• Industry reports and white papers
• Competitor analysis
• Market research data
• Relevant news articles 
• Target keywords
• SEO best practices
• Content gaps analysis 
• Support ticket logs
• Common issues and solutions
• Customer feedback 
• Employee handbooks
• Training materials
• Company policies and procedures 
• Terms of service
• Privacy policies
• Industry-specific regulations 
• Past blog posts
• Newsletters
• Webinar transcripts
• Infographics 
• Videos

Markdown Syntax (.md) is universally understood, very light, and easy to retrieve information from. We transform all of our knowledge base documents into this format.

If it sounds complicated, it's not. All you need to do is format your content onto a Google Doc, click "File" → "Download as .md." Done.

Work well and are easy to edit, but they're heavier files that can slow down processing. AI can read them, but markdown is cleaner and faster. If you're already working in Google Docs, just download as .md when you're done.

Best practice: Copy content into a clean Google Doc, remove unnecessary formatting, then download as .md file.

Text.txt is light but lacks the formatting, leaving it to rely solely on the comprehension of text, symbols, and emojis which makes it difficult to retrieve info at a glance. Plain text is not ideal.

Best practice: Copy content into Google Docs, add clear headings and formatting, then download as .md file.

PDFs can be tricky for AI to process. While AI can read text-based PDFs, the formatting often gets jumbled - tables break apart, columns merge together, and headers get separated from their content.

Some PDFs are actually saved as images (like scanned documents), which makes it more difficult. If you can't highlight and copy text from the PDF, it's an image file.

Best practice: Copy the content into Google Docs, clean up the formatting, then download as .md file.

Common for business presentations, but AI struggles with slide layouts, bullet animations, and design elements. The text is there, but buried in formatting.

Best practice: Copy the content from each slide into Google Docs, organize it logically, then download as .md file. Save any important charts, graphs, or illustrations as separate .PNG files with descriptive names.

CSV and JSON are both excellent formats for structured, tabular data, making them easy for large language models (LLMs) to process and understand.

Excel (.xlsx) can be processed, but the complexity may vary. It is better to convert to CSV for more straightforward data.

Best practice: Save important spreadsheets as CSV files and include them in your folder alongside your .md files.

While many LLMs can access websites, they often struggle with complex page layouts and can be inconsistent when following links.

Website pages also contain ads, navigation menus, sidebar links, and other irrelevant information that can confuse your agent. Articles often include promotional content and links to unrelated topics that have nothing to do with the information you actually need.

Best practice: Copy only the content you actually need from the webpage into Google Docs, remove all ads, navigation menus, and irrelevant links, then download as .md file.

Images are large and more difficult to process quickly. However, in some scenarios, they are irreplaceable in their purpose. You know what they say.

Best practices when using images:

• Use .PNG format for charts, graphs, and text-heavy images (better quality for AI to read)
• Use .JPG format for photos and complex images (smaller file size)
• Clearly title the files. EX: Photo_Restaurant_Main_Dining_Room_View_From_Bar_Seat
• Include information about the images, or at least acknowledge that there are images in the knowledge base
• Keep the pictures focused, avoiding excessive clutter that detracts from their purpose
• If needed, include a short description as a text overlay on the image

Not all LLMs accept video files like MP4. The files are also large and complex, making them difficult to process quickly. Your best move is to transcribe the video into text.

Best practice: Transcribe video content using transcription tools, paste into Google Docs, then download as .md file.