How to Build a WhatsApp AI Chatbot with Google Gemini

What you'll build

• A bot on your own WhatsApp number that answers incoming messages using Google Gemini.
• Grounded in your knowledge base - it quotes your real prices and policies instead of inventing them.
• With conversation memory ('how much is the second one?' works), photo understanding, business-hours rules, and human takeover.

Step 1 - Get a free Gemini API key

Go to aistudio.google.com/apikey, sign in with any Google account, and create a key. The free tier comfortably covers a small business's daily chat volume with a fast model like gemini-2.0-flash.

Step 2 - Connect your WhatsApp number

Create an account on the platform, open Dashboard → Connection, and scan the QR code with the WhatsApp app - the same pairing used by WhatsApp Web. Use a dedicated business number rather than your personal one.

Step 3 - Configure the bot

Open Dashboard → AI Chatbot and fill in three things:

• Gemini API key - the key from step 1. It's yours, so there's no per-reply markup.
• System prompt - the bot's personality: 'You are the friendly assistant for Athens Leather Store. Be brief, warm, and always answer in the customer's language.'
• Knowledge base - plain bullet points of facts: products, prices, sizes, delivery charges, return policy, address, opening hours.

- We sell handmade leather shoes. Sizes 39-45.
- Classic Oxford: Rs 8,500. Loafers: Rs 7,200. Chelsea boots: Rs 11,000.
- Delivery: Rs 300 nationwide, 3-5 working days. Cash on delivery available.
- Returns within 7 days if unworn, buyer pays return shipping.
- Shop: Shop 12, Liberty Market, Lahore. Open 11am-9pm, closed Friday.

💡 The bot is explicitly instructed to never invent prices or policies that aren't in the knowledge base - if something isn't covered, it says a human will follow up. That single rule prevents most chatbot embarrassments.

Step 4 - Turn on the power features

• Conversation memory - keeps the last ~10 messages per chat so follow-up questions resolve correctly.
• Understand images - customers send a photo of a product or a payment receipt; Gemini's vision looks at it and the bot responds accordingly.
• Business hours - typical setup: the bot answers only outside 9:00-18:00, and your team handles chats live during the day.
• Blacklist - your own numbers, family, and suppliers never get a bot reply.

Step 5 - Test and hand off gracefully

Message your number from another phone: ask a price, send a follow-up, send a photo. Then open Dashboard → Inbox - replying manually from there automatically pauses the AI for that conversation for 6 hours, so bot and humans never talk over each other. A per-chat toggle hands control back whenever you want.

Why this beats building it yourself

The DIY route means running a WhatsApp Web bridge (and keeping the session alive), wiring Gemini's API with conversation state storage, handling media downloads and encryption, plus rate-limiting so your number doesn't get banned. All of that is exactly what the platform already does - your job reduces to writing good knowledge-base bullets.