Platform guide

How to improve AI visibility on a WordPress website

By Arnav Mukherjee, founder of TofuBofu · July 6, 2026

If your site runs on WordPress, you are starting from a good place for AI search, and most guides bury that fact under a pile of plugins to buy. The important thing first: WordPress serves server-rendered HTML by default, which means AI crawlers like GPTBot and ClaudeBot can already read your content without running any JavaScript. That is the hard part, and you get it for free.

So the rest is a short, concrete checklist rather than a rebuild. Here is exactly what to do on WordPress, in order.

Step 1: Confirm you are not blocking AI crawlers

WordPress does not block AI bots by default, but SEO and security plugins sometimes do. Open yoursite.com/robots.txt and look for any Disallow rule that hits GPTBot, ClaudeBot, PerplexityBot, or Google-Extended, or a blanket block. Then check your security plugin (Wordfence, iThemes, and similar) for bot-blocking rules that could catch AI crawlers.

If you want AI engines to cite you, these bots have to be allowed. This is the one step with real downside if you skip it.

Step 2: Add FAQ schema

FAQ schema is the highest-ROI structured data for AI, because it hands engines clean question-and-answer pairs. On WordPress, use a schema plugin like Rank Math or a dedicated schema plugin that outputs FAQPage JSON-LD from a simple form. One honest note: after Google stopped showing FAQ rich results, some plugins removed their FAQ blocks. The schema is still valid and still parsed by AI engines, so choose a plugin that still supports it, or paste the JSON-LD into a Custom HTML block.

Step 3: Use real headings, not styled text

In the block editor, use Heading blocks set to H2 and H3, not a Paragraph block made big and bold. Real heading levels give AI engines the structure they use to lift a clean passage. Phrase those headings the way a buyer would ask the question. If you use a page builder like Elementor, use its heading widget with proper levels and keep the page lean, since heavy builders can bury text in generic containers.

Step 4: Write answer-first, one question per page

Give each important page one specific buyer question and answer it in the first sentence, then support it. A page titled for a real query, your service for a particular industry or need, will be retrieved for that query far more reliably than a broad services page. This is the compounding work, and WordPress makes publishing it fast.

Step 5: Verify and build proof

Run the view-source test on a couple of pages to confirm your content and schema are in the HTML. A heavy theme or builder plus aggressive caching can occasionally interfere, so it is worth a look. Then work outside WordPress on the thing WordPress cannot do for you: genuine third-party proof, a few real reviews on the platforms your category uses, which is what AI engines corroborate against.

The exact FAQ schema to paste

If the abstract talk of schema has you unsure what it actually looks like, here is a minimal, valid FAQPage block. Swap in a real question and answer your buyers ask, then on WordPress drop it into a Custom HTML block on the page, or let a schema plugin like Rank Math generate the same output from a form.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "Do you support HIPAA compliance?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "Yes. We provide HIPAA-compliant backups, access controls, and audit logging."
    }
  }]
}
</script>

Add three to six real question-and-answer pairs to the mainEntity array, keep every answer specific and self-contained so it reads well on its own, and after publishing, validate the live page with Google’s Rich Results Test. That one block is the highest-return piece of markup you can ship.

What good looks like: a before and after

Let me make this concrete, because "add schema and structure" stays abstract until you see it play out. Picture a managed IT firm on WordPress. Their services page is one long, handsome block of prose under a hero image: "We deliver comprehensive, reliable IT support tailored to your business." It ranks decently on Google for their brand name. Ask ChatGPT "who should I hire for managed IT for a dental practice in our area," and the firm is nowhere. Not because it is bad. Because there is nothing on that page a machine can lift and attribute with confidence.

Now the after. They keep the same WordPress site and change four things. They split that services page into specific pages, one of which is titled "Managed IT for dental practices," opening with a direct answer: "We provide HIPAA-compliant IT support for dental practices, including practice-management software support, imaging backups, and same-day onsite response." They add an FAQ block with five real questions a dental office asks, marked up with FAQPage schema. They fix the headings so the page uses real H2s that mirror those questions. And they ask three clients for a Clutch review that names the specific work.

None of that touched the platform. It was content and settings inside WordPress. But now, when an engine assembles an answer to that dental-practice question, there is a clean, specific, corroborated passage to quote, and the firm's name is attached to it. That is the whole game, and WordPress makes every step of it a normal afternoon of work.

The mistakes I see most on WordPress

WordPress rarely fails you on its own. It is usually one of a handful of self-inflicted issues, and they are all quick to fix once you know to look.

How to know it worked

The mistake here is treating AEO like a one-time setup. It is a loop, and WordPress's fast publishing is what makes the loop cheap to run. After you ship a batch of changes, do three things. Re-run the view-source test on the pages you touched, to confirm your new content and schema are actually in the HTML. Validate the schema with Google's Rich Results Test so you know it parses. And re-scan your brand across the AI engines to see whether the specific queries you targeted have started naming you.

Expect a lag. Engines re-crawl on their own schedule, so a change made today may show up in answers over the following weeks, not the same afternoon. What you are watching for is movement on the specific questions you wrote pages for, not your overall traffic. When a query that used to name three competitors starts including you, that is the signal that the loop is working, and the cue to point the next page at the next gap.