Upload FAQ

Upload a structured FAQ file and get 40% better answer accuracy on common customer questions.

~10 minutes Free

Why a dedicated FAQ source?

When visitors ask common questions, a dedicated FAQ file consistently produces more accurate and concise answers than crawled web pages. The explicit question-answer structure tells the retrieval system exactly what content maps to which questions.

Bots trained with an FAQ source alongside their website typically show 30–40% fewer incorrect or off-topic answers on the questions covered by the FAQ.


File formats

DGbot accepts FAQ data in two formats:

CSV format:

csv
question,answer
What are your shipping times?,We ship within 2 business days. Standard shipping takes 5–7 days.
Do you offer refunds?,Yes — full refunds within 30 days of purchase. Contact support@example.com to start a return.

JSON format:

json
[
  {
    "question": "What are your shipping times?",
    "answer": "We ship within 2 business days. Standard shipping takes 5–7 days."
  },
  {
    "question": "Do you offer refunds?",
    "answer": "Yes — full refunds within 30 days of purchase. Contact support@example.com to start a return."
  }
]

Both formats produce identical results. Use whichever is easier to maintain.


Steps

1
Prepare your FAQ file

Create a CSV or JSON file following the format above. Each row or object must have a question field and an answer field. No other fields are required.

Start with 20–50 entries
A focused FAQ of 20–50 common questions outperforms an exhaustive one with 500 low-quality entries. Quality beats quantity.
2
Open Sources and click Add source

Go to SourcesAdd sourceFAQ.

Open Sources in admin
training-data/faq-type-picker Screenshot needed — save as: _assets/images/training-data/faq-type-picker.png

Select FAQ from the source type picker.

3
Upload your FAQ file

Drag your CSV or JSON file onto the upload area, or click Choose file to browse.

training-data/faq-upload-form Screenshot needed — save as: _assets/images/training-data/faq-upload-form.png

Upload your FAQ file. DGbot detects the format automatically.

4
Review the parsed entries and confirm

DGbot shows a preview of the parsed Q&A pairs before training. Check that questions and answers look correct.

training-data/faq-preview Screenshot needed — save as: _assets/images/training-data/faq-preview.png

Preview of parsed FAQ entries. Review before clicking Start training.

5
Click Start training and wait for Ready

Click Start training. FAQ sources typically process in under 30 seconds.

training-data/faq-ready Screenshot needed — save as: _assets/images/training-data/faq-ready.png

The FAQ source shows a Ready badge. The bot now prioritises these answers for matching questions.


Writing good FAQ entries

Write questions as visitors actually ask them. "How do I return an item?" beats "Return policy information". Use natural language, not business language.

Keep answers concise. Aim for 1–3 sentences per answer. Longer answers are harder for the bot to use effectively. If an answer needs detail, link to a longer page.

Cover variations. "What's your refund policy?", "Can I return this?", "How do I get my money back?" are the same question asked three ways. Add all three as separate entries with the same answer.

Update when your policies change. The FAQ is often the highest-impact source. An outdated FAQ answer misleads visitors more than an outdated web page does.