Respond to WhatsApp messages with AI like a pro

WhatsApp message → n8n type-based routing (text/image/audio/video) → Gemini multimodal understanding → AI Agent generates reply → WhatsApp send-back

Who

Users who want to build a WhatsApp AI chatbot in n8n

You get

Auto-receive and understand WhatsApp text, image, voice, and video messages, generate replies via AI Agent, and send them back

You need

n8n · whatsapp · gemini

Steps

1. Configure WhatsApp trigger and message routing

   After importing the template into n8n, configure the WhatsApp Trigger node (subscribe to messages updates) and set the Split Out Message Parts node to split the messages array. Then use the Switch node (Redirect Message Types) to route by type: audio, video, image, and text (fallback).

2. Transcribe audio messages

   For the audio branch: use the WhatsApp node to get the audio file URL (Get Audio URL), then download the audio via HTTP Request. Call the Google Gemini API (via HTTP Request) to transcribe the audio content.

   Execution prompt · Source text。Google Gemini Audio node JSON Body field

   Copy prompt

   ={{
   {
     "contents": [{
       "parts":[
         {"text": "Transcribe this audio"},
         {"inlineData": {
           "mimeType": `audio/${$binary.data.fileExtension}`,
           "data": $input.item.binary.data.data }
         }
       ]
     }]
   }
   }}

3. Describe video messages

   For the video branch: use the WhatsApp node to get the video file URL (Get Video URL), download via HTTP Request. Since n8n's LLM node does not currently support video binary types, call the Google Gemini API directly via HTTP Request to describe the video.

   Execution prompt · Source text。Google Gemini Video node JSON Body field

   Copy prompt

   ={{
   {
     "contents": [{
       "parts":[
         {"text": "Describe this video"},
         {"inlineData": {
           "mimeType": `video/${$binary.data.fileExtension}`,
           "data": $input.item.binary.data.data }
         }
       ]
     }]
   }
   }}

4. Analyse image messages

   For the image branch: use the WhatsApp node to get the image URL (Get Image URL), download it, then pass to the Image Explainer (Basic LLM Chain) node where the Gemini model describes the image and transcribes visible text.

   Execution prompt · Source text。Image Explainer node Text field

   Copy prompt

   Here is an image sent by the user. Describe the image and transcribe any text visible in the image.

5. Summarize text messages

   For the text branch (fallback): pass through the Wait node (Get Text) to the Text Summarizer (Basic LLM Chain) to succinctly summarize the user's message for the AI Agent.

   Execution prompt · Source text。Text Summarizer node Messages > Message field

   Copy prompt

   Summarize the user's message succinctly.

6. Generate response with AI Agent

   Results from all branches converge at the Get User's Message node, extracting message_type, message_text, from, and message_caption. This feeds into the AI Agent node, which uses Window Buffer Memory (session keyed by phone number) and the Wikipedia tool to generate a reply.

   Execution prompt · Source text。AI Agent node Text field

   Copy prompt

   =The user sent the following message
   message type: {{ $json.message_type }}
   message text or description:
   ```{{ $json.message_text }}```
   {{ $json.message_caption ? `message caption: ${$json.message_caption.trim()}` : '' }}

   Execution prompt · Source text。AI Agent node Options > System Message field

   Copy prompt

   You are a general knowledge assistant made available to the public via whatsapp. Help answer the user's query succiently and factually.

7. Respond to WhatsApp user

   The AI Agent output is sent back to the original sender via the WhatsApp node (Respond to User). This node supports text, images, audio, video, documents, and location; this template demonstrates text-only replies.

You are an n8n workflow setup assistant. The user wants to replicate the 'WhatsApp AI Chatbot' template. Assist in this order: 1) Confirm the user has an n8n instance, WhatsApp Business account, and Google Gemini API credentials; 2) Guide importing workflow.json and configuring the three credentials (whatsAppTriggerApi, whatsAppApi, googlePalmApi); 3) Explain WhatsApp webhook setup before activating the workflow; 4) Warn that large media files may impact performance and suggest pre-processing.