I’m fairly new to working with n8n but really enjoying the experience. I’ve managed to get several workflows running using different templates and examples. Now I want to create a chatbot that can pull answers from my WordPress site’s FAQ section.
My current configuration:
WordPress node is connected and authentication works fine
Resource type is set to Page with Get operation
FAQ page ID is properly configured
The problem I’m facing:
When users ask questions, I can see the WordPress node is active and processing, but the chatbot keeps giving ChatGPT responses instead of actually using the content from my FAQ page.
How can I make sure the system checks my FAQ content first before using ChatGPT as a backup? Any suggestions for optimizing this workflow would be really helpful.
Check if your ChatGPT node has higher priority than WordPress in the execution order. Sounds like they’re both firing simultaneously instead of one after the other. Try adding a merge node between them - that’ll make ChatGPT wait for WordPress to finish processing first.
This sounds like a workflow execution and conditional logic issue. I ran into the same thing building my customer support bot last year. You need proper branching logic between your WordPress and ChatGPT nodes. Here’s what fixed it for me: Add an IF node after WordPress content retrieval to check if relevant FAQ content was found. Set up conditions that verify the retrieved content has keywords matching the user’s query. If the content score hits your threshold, send it to a response formatter. If not, trigger ChatGPT as fallback. Also crucial - preprocess the user input before the WordPress lookup. I added a simple text processing step to extract key terms from questions. This massively improved matching accuracy against FAQ content. Without it, you’re probably pulling the entire FAQ page but not matching it to the specific question. Double-check your workflow execution settings have the right node priorities too. Otherwise ChatGPT might run simultaneously instead of being a true fallback.
Had the same setup for 6 months - you’re missing content scoring. Your WordPress node grabs the full FAQ page but can’t tell if it actually answers the user’s question. I fixed this by adding a similarity check between user input and individual FAQ entries. Split the FAQ content into separate question-answer pairs, then score each pair against the user’s query with basic text matching. Set a threshold - I use 70% similarity. Hit that mark? Return the FAQ answer. Miss it? Fire up ChatGPT. Also, check your workflow execution order. Make sure WordPress finishes completely before ChatGPT starts. Running them simultaneously might trigger ChatGPT while WordPress is still working. This dropped my ChatGPT usage by 60% since most questions are already in the FAQ.
Your WordPress node’s pulling the FAQ page fine, but you’re missing the content filtering and search logic. I ran into this exact issue with a client’s knowledge base. Pulling the whole FAQ page won’t cut it - you need keyword extraction and content chunking. Here’s what worked for me: add a Code node after the WordPress retrieval that splits FAQ content into individual entries, then use basic string matching to find relevant sections. The key part is proper error handling. Only pass control to ChatGPT when WordPress returns empty or irrelevant results. I’m betting your setup has both nodes running at the same time instead of following a real conditional path. Toss in some debug logging to see what content WordPress is actually returning - you’ll probably find it’s grabbing the page but not the specific FAQ answers you’re after.
your workflow isn’t parsing the FAQ content right. i ran into this too - the WordPress node pulls the entire page, but you’ve got to extract the specific answers. add a text processing step after the WordPress retrieval to hunt for keywords that match the user’s question. and make sure ChatGPT only kicks in when there’s no FAQ match.
This happens because n8n runs parallel paths without checking what actually got returned. Your WordPress node grabs the FAQ content fine, but ChatGPT fires anyway regardless of what WordPress found. I hit the same thing building a docs bot. Here’s what fixed it: Add a function node right after WordPress that checks if the FAQ content actually answers the user’s question. I use simple keyword matching or check how relevant the content is with a basic scoring system. Only trigger ChatGPT when that relevance score is too low. Key thing - don’t pull the entire WordPress page. Target specific FAQ sections and break them into individual Q&A pairs before matching against what the user asked. This cut my ChatGPT usage way down while keeping responses just as accurate.
check your node order - chatgpt’s probably firing before wordpress finishes processing. drop a ‘wait’ node after the wordpress retrieval so it has time to wrap up. also, wordpress spits out raw html, so you’ll want to strip those tags before trying to match content.