How to Upload Documents to Train Your AI
Before You Start
Make sure your document is in a text-based format: PDF, TXT, or DOCX. The file should contain actual text, not scanned images. If you have a scanned document, run it through OCR first. Review the content for accuracy, since the chatbot will treat everything you upload as authoritative information.
If your document is very long (over 50 pages), consider splitting it into sections by topic. This helps the chunking process create more focused, relevant pieces. A 100-page manual split into 10 topic-specific documents will produce better search results than the same content uploaded as one massive file.
Step-by-Step Upload Process
Log into your admin panel at admin.aiappsapi.com. Click on the AI Chatbot app from your dashboard. If you have not installed it yet, find it in the app directory and click Install.
Choose the chatbot you want to add knowledge to, or create a new one by clicking Create New Chatbot. Each chatbot has its own knowledge base, so information uploaded to one chatbot is not automatically shared with others.
In the chatbot's edit page, find the knowledge base or embeddings section. This is where all training data for this chatbot is managed.
You have two options. Click the file upload button to select a PDF, TXT, or DOCX file from your computer. Or use the text input area to paste content directly. Pasting is useful when you want to add a specific Q&A list, a product description, or information you have written specifically for the chatbot.
Adding a tag or label to your upload helps you organize and manage your training data later. For example, tag product information as "products," policies as "policies," and FAQ content as "faq." This makes it easy to find and update specific content later.
Click the upload or submit button. The system will chunk your content into pieces and generate embeddings for each chunk. Processing typically takes 5 to 30 seconds depending on document length. You will see the new entries appear in your knowledge base list when processing is complete.
Open the chatbot's test interface and ask questions that should be answered by the content you just uploaded. Verify the responses are accurate and reference the correct information. If the chatbot misses something, check that the relevant content was included in your upload.
Tips for Better Results
- Use clear, descriptive text. Write as if explaining to a customer. Avoid internal jargon unless your chatbot serves an internal audience that uses that jargon.
- One topic per section. If your document covers multiple topics, make sure each section is self-contained. The chunking process works best when each chunk is about one thing.
- Include context. Instead of just listing a price like "$29.99," write "The Basic plan costs $29.99 per month and includes..." This gives the embedding more context to work with.
- Remove boilerplate. Strip out repeated headers, footers, legal disclaimers, and navigation text that appears on every page. This content adds noise without adding useful knowledge.
Managing Uploaded Content
After uploading, you can view all chunks in the knowledge base list. You can delete individual chunks or entire uploads if the information becomes outdated. To update content, delete the old version and upload the new one. The chatbot immediately stops using deleted content and starts using new content.
There is no limit on how many documents you can upload to a single chatbot. Add content as needed, and the chatbot's knowledge grows with each upload. See How to Keep Your AI Training Data Up to Date for ongoing maintenance tips.
Upload your first document and watch your chatbot start answering questions with your own business knowledge.
Get Started Free