How to Upload your PDF Catalog to a WhatsApp Chatbot

If you have a business and already have a product catalog in PDF format, you've probably wondered: "Can I have my WhatsApp chatbot use this catalog to answer questions from my customers?" The answer is yes, and in this tutorial we explain exactly how to do it.

Uploading your PDF catalog to a WhatsApp chatbot is one of the fastest ways to train artificial intelligence to know your products, prices, descriptions and availability. Instead of uploading product by product manually, you simply upload the file you already have and the system takes care of the rest.

Why Upload your PDF Catalog to a Chatbot?

Before we get into the technical steps, let's look at why this strategy is so effective for businesses of all sizes.

You have the file ready

Most businesses already have a PDF catalog. Whether you use it to email, share on social media or print for your customers, that same file can become the knowledge base for your chatbot. No need to create new content from scratch.

Quick and easy configuration

While uploading products one by one into a structured database can take hours or even days depending on the size of your catalog, uploading a PDF takes literally minutes. It's the fastest way to get your chatbot up and running with real product information.

AI can answer specific questions

Thanks to RAG (Retrieval-Augmented Generation) technology, artificial intelligence not only stores your catalog, but also understands it. This means that when a customer asks "Do you have sneakers in size 42?", the chatbot can search your catalog and give an accurate answer based on the available information.

Reduce your team's workload

Every question the chatbot answers automatically is one less query for your sales or customer service team. This allows them to focus on closing sales and handling more complex cases while the chatbot handles repetitive product inquiries.

Step 1: Prepare your PDF for Maximum Readability by AI

Not all PDFs are equal when it comes to being processed by artificial intelligence. A well-structured PDF will generate better chatbot responses. Here's how to optimize your file.

Use real text, not images of text.

This is the most important point. If your catalog is a PDF created from scans or screenshots, the AI will have difficulty reading the content. Make sure that the text in your PDF is selectable, i.e. you can copy and paste it. If you cannot select the text, you will need to recreate the catalog in a digital format or use OCR tools first.

Clear structure by product

Each product should have a clearly identifiable section. Ideally, you should include:

• Product name in bold or highlighted text
• Code or SKU if you handle it
• Price clearly visible next to the product
• Description detailed with main features
• Availability or variants (colors, sizes, etc.)

Use tables whenever possible

Tables are excellent for AI to identify the relationship between a product and its attributes. For example:

Product	Price	Available Sizes	Color
Runner Pro Shoe	$89.99	38, 39, 40, 41, 42	Black, White
Comfort Sandal	$45.99	36, 37, 38, 39	Beige, Brown

This format makes it much easier for the chatbot to extract accurate information when a customer asks a specific question.

Avoid overly complex designs

Catalogs with many overlapping columns, rotated text, complex graphic elements or text over images can confuse the document processor. Maintaining a clean, linear layout significantly improves the quality of responses.

Includes headings and categories

Organize your catalog by clear sections: "Sport Shoes", "Casual Shoes", "Accessories", etc. Headings help the AI understand the context and correctly categorize each product.

Step 2: Upload the PDF to the Platform

Once your PDF is optimized, the next step is to upload it to your chatbot platform. At Aurora Inbox, this process is very simple:

1. Login to your administration panel and log in with your credentials.
2. Navigate to Knowledge Base section o "AI Training".
3. Click on "Upload Document". and select your PDF file.
4. Assign a descriptive name to the document, for example: "Spring 2025 Catalog - Footwear".
5. Confirms the load and wait for the system to process the file.

The maximum file size may vary by platform, but PDFs up to 50 MB are generally acceptable. If your catalog is very large, consider dividing it into thematic sections to facilitate processing.

Step 3: AI Indexes the Content

This step occurs automatically after uploading your file. The platform uses document processing technology to:

1. Extract all text of the PDF, including tables and lists.
2. Split the content into fragments logical and manageable.
3. Create vector embeddings of each fragment for semantic search.
4. Store in the knowledge base for future reference.

This process can take anywhere from a few seconds for small catalogs to several minutes for large documents. Once completed, you will see a confirmation that the document was successfully indexed.

What is semantic search?

Unlike an exact keyword search, semantic search understands the meaning behind the question. If a customer asks "What options do you have for running?", the AI will understand that they are referring to running shoes, even if those exact words do not appear in your catalog.

Step 4: Test with Product Questions

Before putting your chatbot into production, it is essential to test it with different types of questions that your real customers might ask. Here are some examples to validate:

Pricing questions

• "How much does the Runner Pro Sneaker cost?"
• "What is the cheapest product you have?"
• "Do you have anything for less than $50?"

Availability questions

• "Does the Comfort sandal come in size 38?"
• "What colors are available for the Runner model?"
• "Do you have products in red?"

Characteristic questions

• "What is the difference between the Runner model and the Sprint model?"
• "What materials does the Comfort sandal use?"
• "Which shoe do you recommend for long distance running?"

General questions

• "What products do you have?"
• "Show me your catalog of sports cars."
• "Do you have any promotions or discounts?"

Write down the questions where the chatbot does not answer correctly or gives incomplete information. This will help you in the next step.

Step 5: Refine and Improve

Based on the tests in the previous step, you can improve the results in several ways:

Update PDF

If you detect that certain questions are not answered well because the information is not clear in the document, edit your PDF to make that information more explicit. For example, if customers frequently ask about return policies and that is not in your catalog, add it.

Add complementary documents

You can upload multiple documents to enrich the knowledge base:

• Size guide with detailed measurements
• Shipping and return policies
• Frequently Asked Questions about your products
• Technical Data Sheets of specific products

Configure customized responses

For very specific or critical questions for your business, some platforms allow you to set up manual responses that take priority over searching the catalog. This is useful for frequently changing information such as current promotions.

Monitor conversations

Regularly review chatbot conversations to identify patterns of questions that are not answered correctly and adjust your knowledge base accordingly.

Best Practices for PDF Formatting

To get the best results with your chatbot, follow these formatting recommendations:

Recommended structure by product

PRODUCT NAME
Code: SKU-12345
Price: $XX.XX
Description: Brief description of the product with its main features.
Material: Synthetic leather / Fabric / etc.
Available sizes: 36, 37, 38, 38, 39, 40
Colors: Black, White, Blue

Additional formatting tips

• Uses standard fonts such as Arial, Helvetica, or Times New Roman
• Maintains a readable font sizeminimum 10pt
• Avoid colored backgrounds behind the text
• Do not use watermarks that are superimposed on the product text
• Includes an index at the beginning if the catalog is more than 20 pages long
• Number the pages for easy reference
• Clearly separate sections with page breaks or dividing lines

Limitations of the PDF Catalog vs. the Structured Catalog

It is important to be transparent about the limitations of using a PDF as a knowledge base versus a structured catalog:

Limitations of PDF

• Not updated in real time: If you change a price or a product is out of stock, you must upload a new PDF.
• Variable precisionAI interprets the text, which may generate occasional inaccuracies in numerical data.
• No inventory management: The chatbot cannot check stock in real time.
• Dependent formatIf the PDF has a complex format, the data extraction may be imperfect.

Advantages of the structured catalog

A structured catalog (loaded product by product with defined fields) offers:

• Individual product updates
• Accurate pricing and availability
• Integration with inventory systems
• More accurate search and filtering

Which one to choose?

The recommendation is to start with the PDF to get your chatbot up and running quickly, and gradually migrate to a structured catalog if your volume of products and queries justifies it. Many businesses use both: the structured catalog for core products and supplemental PDFs for additional information.

How Aurora Inbox Does It

Aurora Inbox offers a complete solution to upload your PDF catalog and train your WhatsApp chatbot:

PDF document upload

From the administration panel you can upload one or multiple PDF documents. The system accepts catalogs, manuals, price lists and any document containing relevant information for your customers.

Advanced RAG technology

Aurora Inbox uses Retrieval-Augmented Generation (RAG) technology to process your documents. This means that the AI does not memorize your catalog, but actively searches for relevant information every time a customer asks a question, ensuring accurate and contextual responses.

Continuous training

You can update your knowledge base at any time by uploading new versions of your catalog. The system will automatically reindex the content to keep the answers up to date.

Direct integration with WhatsApp

The entire process is natively integrated with WhatsApp Business, which means your customers can ask questions about products directly in their WhatsApp chat and receive instant answers based on your catalog.

Conclusion

Uploading your PDF catalog to a WhatsApp chatbot is the fastest and most affordable way to automate product support. With proper document preparation and following the steps in this tutorial, you can have your chatbot answering questions about your catalog in minutes.

Remember that the key is the quality of the PDF: a well-structured document with clear text, organized tables and defined sections will produce a much more effective chatbot. And if over time you need more precision, you can always complement it with a structured catalog.

Ready to get started? Try Aurora Inbox and upload your first catalog today to see how your WhatsApp chatbot becomes a 24/7 sales assistant.

---

Frequent questions

Which PDF format works best for the chatbot?

Digitally generated PDFs (from Word, Google Docs, InDesign, etc.) perform significantly better than scanned PDFs. Text must be selectable and copyable. If your catalog is a scan, first convert it to digital text using OCR (Optical Character Recognition) tools before uploading it to the chatbot.

Can I upload several PDFs or just one?

You can upload multiple PDF documents to your knowledge base. In fact, it is advisable to separate the information into thematic documents (product catalog, shipping policies, size guide, etc.) so that the chatbot can locate the information more efficiently. In Aurora Inbox there is no limit to the number of documents you can upload.

What happens if I update my product catalog?

When your products, prices or availability change, simply upload the new version of the PDF. The system will reprocess the document and update the knowledge base automatically. We recommend you to delete the previous version to avoid the chatbot using outdated information and generate confusion with your customers.

Can the chatbot show images of the products in the PDF?

Currently, most RAG-based chatbot systems extract and use the text from the PDF, not the images. The chatbot can describe the product in detail based on the text, but to display images you would need a structured catalog with links to photos of each product. Some platforms are integrating multimodal capabilities that could change this in the near future.

How long does it take for the chatbot to be ready after uploading the PDF?

Processing time depends on the size of the document. A 10-20 page catalog is usually processed in less than a minute. Larger catalogs of 100+ pages can take 3-10 minutes. Once processed, the chatbot is immediately ready to answer questions based on the content of the document. In Aurora Inbox you will receive a notification when the indexing process is complete.