From visitor questions to a smarter content plan: An OpenClaw experiment

What if your WordPress site could help visitors find the content they need, and help you figure out what to write about next?

The secret to a successful WordPress blog is creating content that’s highly useful and relevant to your readers.

But how do you know what your readers really want to know? (After all, you’re not a mind reader!)

Your website can tell you which pages people visited, but it can’t tell you what people came looking for and failed to find.

What if a visitor with an unanswered question could be the starting point for a useful new article, a better page or a smarter internal link?

As we mentioned in our OpenClaw announcement, we’ve been playing around with creative ways for WordPress site owners to use this always-on AI agent.

So, we tried a simple idea: Could an AI assistant help visitors find what they need from your content, and also help you improve your website?

The result was a working WordPress experiment we called Site Brain.

It’s a practical example of what you can build with OpenClaw by combining it with WordPress and a few supporting services. We’ll show you how the pieces come together, so you can experiment with building something similar yourself.

What is Site Brain? How Does it Work?

In our experiment, Site Brain took the form of a simple, user-friendly AI chatbot in the corner of a WordPress site, ready to help visitors explore its content.

In a nutshell:

1. First, Site Brain answers questions by directing visitors to relevant content.
2. Then, unanswered questions become blog post ideas.
3. Blog post ideas could become your next editorial plan.
4. Site Brain could also spot your orphan posts…
5. And create a plan for rescuing them.

Easy-to-use and intuitive, the experiment shows how an assistant like this could turn a static site into an interactive resource your visitors will want to come back to again and again.

First, Site Brain answers questions by directing visitors to relevant content.

We tested the prototype on a food blog. You can explore the demo here to see the concept in action.

The chat became a “Recipe Helper,” where the user could ask questions naturally, just like they would interact with any chatbot.

For example:

• What can I cook with okra?
• Show me some quick vegetarian dinners.
• What are ideas for breakfast recipes that use bananas?
• I’m craving something spicy. What should I cook tonight?

The assistant then searches the existing recipes and returns relevant suggestions, with links to real posts.

Rather than being a replacement for your content, the assistant directs readers to the exact blog posts they need, helping your site visitors discover and benefit from as many of your posts as possible.

Then, unanswered questions become blog post ideas.

Behind the scenes, the agent reviewed the questions visitors asked and checked how well the site’s content answered them. It reported common questions, the overall answered rate and any questions the site’s content didn’t answer particularly well.

Before Site Brain, visitors would simply bounce if your content didn’t answer their question – and you’d never know why.

With an assistant like this, you could have a source of possible content ideas based on something a real visitor wanted to know when they came to your site.

When it comes to content creation, that knowledge is absolute gold.

Blog post ideas could become your next editorial plan.

In the demo, visitors asked about quick vegetarian dinners, Indian food, potatoes, tiramisu and diet-friendly suggestions.

The agent compared these questions with the site’s existing recipes and suggested five possible articles. Each idea is connected to actual visitor demand, fills a gap in your content and connects with relevant content already on the site.

Now, rather than starting from scratch every time or guessing what to publish next, this could give a site owner an editorial calendar full of ideas based on real visitor demand.

Site Brain could also spot your orphan posts…

When a website page has no other pages linking to it, it’s known as an “orphan.”

The page might have really valuable or useful content, but visitors are unlikely to discover it and search engines receive very few signals about how it fits into the rest of the site. The searchable memory created by Site Brain could help to address this problem.

The private agent builds a picture of the site’s internal links and identifies pages with no inbound links, so you can build a path to them from the rest of your content.

In the demo, for example, the agent found that the Baingan Bharta recipe had nothing else on the site pointing to it.

And create a plan for rescuing them.

Identifying the orphan post was only the first step. In our demo, the agent then offered to create an internal link plan to connect this post with other relevant content.

Rather than inventing anchor text, the agent looked for words that already appeared in related recipes and quoted the exact sentence where each link would be a good fit.

This makes every suggestion super easy for you to review. The prototype can preview a proposed fix and wait for confirmation before applying it. It even saves a snapshot first, allowing the change to be reversed.

At every stage of the process, you’re still in control. In our prototype, the agent wasn’t set up to write or publish blog posts on the user’s behalf. Instead, it took care of the repetitive discovery work behind the scenes, while leaving every proposed change for the site owner to review and approve.

In our prototype, the index was deliberately configured to read only public posts, not drafts, private posts or password-protected content.

How the Site Brain prototype works

Our prototype has three pieces that work together:

• A WordPress plugin reads your site’s public content and keeps a searchable index up to date.
• A visitor-facing OpenClaw assistant uses that index to answer questions and link to the site’s real pages.
• A private team agent uses the same information to find content gaps, orphan pages and internal-link opportunities.

OpenAI embeddings make the content searchable, while Qdrant stores the index. The search understands meaning, so it’ll connect a visitor’s question with relevant pages even if they use different wording.

But don’t worry, you don’t need to understand all the technical details to see why this idea is so exciting!

Site Brain brings together OpenClaw, WordPress, OpenAI embeddings, and Qdrant to create the experience you’ve seen above.

There are a few moving parts to configure, so we’ve put together a step-by-step guide to help you set it up and experiment with building your own version.

Think of Site Brain as a working experiment and a starting point. You can adapt it to your own site, change how it works, or take the idea in completely different directions, all with OpenClaw.

Site Brain helps both you and your visitors get more out of your content

The experiment demonstrates a potentially powerful loop:

1. A visitor asks a question.
2. The assistant helps them using your site’s existing content.
3. The private agent learns where the content was useful and where it fell short.
4. As the site owner, you get practical ideas for what to write, update, or link next.

Instead of a chat widget ending when the conversation ends, it becomes the beginning of a better content plan.

The experiment was tested on a food blog in this example, but the same idea works well for any content-heavy website. For example, this could be used for a database of technical tutorials, a travel blog with practical destination guides, or a library of educational resources.

The Site Brain experiment shows how an OpenClaw-powered assistant could help you build a stronger connection with your readers and a deeper understanding of what they’re looking for when they arrive on your site. With this, you can answer their questions better today, and create even more helpful content for the next time they visit.

This is just one of the many cool things that are possible when you run OpenClaw on WPMU DEV hosting!

What do you think? Could you imagine building or using something like this on your site? We’d love to hear your thoughts in the comments.