Every AI chatbot plugin promises answers based on your site, not generic web noise. Few explain how that happens, and even fewer let you do it for free. This guide gives a clear, non-technical explanation of the technique behind those claims — RAG — then walks through setting up Hyve Lite, a free WordPress plugin that answers visitors using your posts and pages.
Quick disclosure: Hyve is our plugin from Themeisle (the team behind WPShout). Expect straightforward, honest notes about limitations and what the free version doesn’t include.
How RAG (retrieval-augmented generation) works — in three simple steps
1) Your question becomes a number. When a user asks something, the chatbot converts that question into an embedding: a numeric fingerprint representing its meaning.
2) Your content is searched by meaning. Your posts and pages are already split into small text chunks and converted to embeddings. The system finds the chunks whose embeddings are closest to the question’s embedding.
3) The AI writes a grounded answer. The retrieved chunks plus the original question go to the model, which composes an answer grounded in your text.
Practical consequences: the good and the realistic
– Good: Answers reflect what you actually published. Update a post, re-sync, and the bot’s replies change.
– Realistic: The quality of answers depends on the quality of your content. Thin content yields thin answers.
Setting up a WordPress AI chatbot on your content (about 15 minutes)
Step 1 — Install Hyve Lite
Go to Plugins > Add New in your dashboard, search for “Hyve,” then install and activate Hyve Lite. Activation adds a Hyve menu in the sidebar. No coding required.
Step 2 — Choose how the AI runs: your OpenAI key or Hyve Connect (Pro)
– Using your own OpenAI key: In Hyve > Settings > Advanced paste your OpenAI API key and Save. Messages go from your site to OpenAI and consume your account credits.
– Using Hyve Connect (Pro): With a Pro license you skip the API key. Connect via the Hyve Connect panel; Hyve handles the model and billing and shows pooled usage meters.
Step 3 — Build the knowledge base
Open Hyve > Knowledge Base and click Add Data. You’ll see your posts and pages with an Add button next to each title. Choose what should power answers. Don’t add everything — pick the most informative and relevant items (FAQs, docs, explainers, policies). The selection determines how well the assistant responds.
Step 4 — Tune the assistant
Under Hyve > Settings > Assistant choose the model and sampling behavior. Options run from GPT-4.1 down to GPT-4o mini. More capable models cost more per message; GPT-4o mini is a sensible starting point. Use the Temperature control to set randomness: lower values make replies focused and consistent (recommended for support-style bots), higher values make them more creative.
Step 5 — Turn the chat on
From Hyve > Dashboard enable chat on all pages or add it selectively with the Hyve Gutenberg block or shortcode. Then test on the front end by asking something that should be answered from a published post. If the bot quotes your text, the RAG pipeline is working.
Key trade-offs and final thoughts
Hyve Lite wraps the RAG pipeline in an easy-to-use free plugin. Important trade-offs: you still pay for the AI (either from your OpenAI account or via Pro’s pooled billing), the free knowledge base is capped (500 chunks), and advanced reporting is a Pro feature. For a blog, documentation site, or small business site with solid content, this is a low-risk way to learn what visitors ask and deliver answers grounded in your own text. If you prefer not to manage an OpenAI account, Hyve Connect (Pro) handles the AI for you.
Try pointing the plugin at your ten best posts and see what questions visitors ask — you may be surprised by what they actually want to know.
Have you tried running a chatbot on your site? What unexpected questions did visitors ask?