Automated Fact-Checking and Evaluation in WordPress Workflows

This n8n workflow is designed to automate the process of evaluating the accuracy and correctness of AI-generated responses against ground truth data. It integrates multiple nodes including language models, data fetchers, and evaluation metrics to streamline fact-checking workflows, especially useful for content moderation, educational platforms, or AI validation scenarios. The workflow begins by fetching dataset rows from a Google Sheet, where each row contains questions, answers, and ground truth data. An n8n webhook triggers the process upon receiving new chat messages or dataset entries. The workflow then leverages OpenAI’s GPT models to generate responses, compute embeddings, and calculate similarity scores. It classifies responses into true positives, false positives, and false negatives, then calculates an overall correctness score via F1 metric. Results are stored back into Google Sheets, providing continuous monitoring and validation of AI responses in real-time. This setup is useful for improving AI training quality, ensuring factual accuracy, and maintaining high standards in automated content or chatbot systems.