See llms.txt for all machine-readable content.

Back to Templates

Validate AI answers with Google search, OpenAI GPT-4, Slack and Gmail alerts

Created by

Created by: WeblineIndia || weblineindia
WeblineIndia

Last update

Last update a day ago

Categories

Share

Quick overview

This workflow receives an AI-generated Q&A via webhook, searches the web for supporting sources, and uses OpenAI to score and classify the answer as SAFE, SUSPICIOUS, or HALLUCINATED, then alerts via Slack or Gmail and returns a response to the webhook caller.

How it works

  1. Receives a POST webhook request containing a question and answer.
  2. Sends the question to a Google Search API endpoint and retrieves the top results.
  3. Formats the top search snippets into a compact source list for validation.
  4. Sends the question, answer, and sources to OpenAI (GPT-4 Turbo) to return a JSON verdict with a confidence score, reasoning, and classification.
  5. Routes the result by classification, returning the result to the webhook caller for SAFE responses.
  6. Sends a Slack message to a chosen channel when the response is classified as SUSPICIOUS.
  7. Sends a critical Gmail alert email when the response is classified as HALLUCINATED.

Setup

  1. Configure the incoming system to POST question and answer to the workflow’s webhook URL at /hallucination-check.
  2. Add OpenAI credentials in n8n and ensure the selected model (GPT-4 Turbo) is available for your account.
  3. Replace the placeholder Search API URL and X-API-KEY header in the HTTP request node with your Google/Serp API provider endpoint and key.
  4. Add Slack credentials, select the target channel, and confirm the workflow can post messages.
  5. Add Gmail credentials and replace your_email_here with the recipient address for critical alerts.

Additional info

Need Help?

If you need help improving this workflow, scaling it or adding advanced features like multi-model validation or AI monitoring dashboards, our n8n developers at WeblineIndia are happy to help and enhance it into a production-grade AI safety system.