Kaliber Group — Performance Marketing, Augmented by AI

How to Fix AI Readability on WordPress (2026)

19 May 2026

Why Does AI Readability Matter for WordPress Sites?

WordPress powers roughly 43% of all websites (W3Techs, 2025: https://w3techs.com/technologies/details/cm-wordpress). The platform's flexibility is part of why it dominates — and also part of why AI Readability scores vary so wildly across WordPress sites. Two sites running the same theme, the same hosting, the same general business model can score 30 points apart in an AI Readability audit purely because of how each was configured at the schema, robots, and content layer.

Most of these gaps are not visible to the site owner. WordPress's admin UI does not surface whether your robots.txt allows ClaudeBot, whether your Organization schema includes a logo block, or whether your headings on the homepage skip from H1 directly to H3. AI engines read all of these signals. Site owners rarely look at any of them.

The fixes that follow are ranked by the weights used in the Kaliber AI Readability Checker, which scores sites against the structural signals AI engines actually consume. The ranking reflects what produces the largest score gain per hour of work — not the order in which they happen to appear in a CMS settings panel.

What Did Google's May 2026 AI Search Guide Actually Say?

On May 7, 2026, Google published an updated AI Search documentation page that consolidated their official stance on AI Overviews, AI Mode, and the broader category of generative AI search. The headline claim: "optimizing for generative AI search is optimizing for the search experience, and thus still SEO" (Google Search Central, 2026: https://developers.google.com/search/docs/appearance/ai-features).

Three points in that guide directly contradict advice circulating in the GEO/AEO community as recently as last quarter:

1. No special files are needed. Google explicitly says "no special machine-readable files or AI-specific markup are necessary." This is a direct rejection of the llms.txt proposal that gained traction in early 2025 — a community-led idea to publish a flat-text file at /llms.txt describing your site's content to large language models. Google says it provides no benefit for their AI search surfaces. It may still help niche LLM parsers that explicitly support it (Cloudflare, some smaller AI tools), but the broad-adoption hypothesis is now contradicted by Google's own guidance.

2. No content chunking is needed. Google states "there's no requirement to break content into small pieces for AI systems." This dismisses the practice of publishing parallel Markdown mirrors at /page.md URLs or breaking long pages into shorter ones for AI parsing. Their systems handle multi-topic pages natively.

3. No special schema is needed for AI. Per the guide: "structured data ... isn't required for generative AI search and there's no special schema.org markup to add." This is an important caveat — schema is still useful for traditional search rich results (where they still apply) and for non-Google LLMs that explicitly consume structured data. But "add schema specifically because of AI search" is not a defensible position with Google.

Separately, Google announced on the same release that FAQ rich results would stop appearing in search effective May 7, 2026, with Search Console reporting removed in June and API support ending August 2026 (Search Engine Journal, 2026: https://www.searchenginejournal.com/google-drops-faq-rich-results-from-search/574429/). The schema itself remains valid — sites can keep it without indexing penalty — but the rich-result visibility is gone.

How Do You Allow AI Crawlers in WordPress robots.txt?

This is the single highest-impact fix in the audit. AI engines cannot read content they cannot fetch. If your robots.txt blocks GPTBot or ClaudeBot — or if it only allows a generic User-agent: * rule that may have been narrowed by another plugin — your site is effectively invisible to those engines.

WordPress by default serves a virtual robots.txt generated dynamically, which means there's no file on disk to edit directly. You either use a plugin to override it or create a physical robots.txt in your site's root via FTP/SSH.

The robots.txt block to add

Add the following block above (or in place of) any existing rules:

User-agent: GPTBot
Allow: /

User-agent: ClaudeBot
Allow: /

User-agent: anthropic-ai
Allow: /

User-agent: CCBot
Allow: /

User-agent: Google-Extended
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: Applebot-Extended
Allow: /

User-agent: cohere-ai
Allow: /

User-agent: Bytespider
Allow: /

Sitemap: https://yourdomain.com/sitemap.xml

If your site has sensitive areas you want to block from AI training while still allowing search indexing, the asymmetry matters. Google-Extended controls Google's training data without affecting their search crawler. Applebot-Extended does the same for Apple Intelligence. Setting Disallow: / on these blocks training without losing search visibility. For most marketing sites, allowing both is the default position.

How Do You Fix Heading Hierarchy in WordPress?

AI engines use heading structure to understand what a page is about and how its subtopics relate. The audit penalises pages that have more than one H1, that skip levels (H1 directly to H3, no H2), or that have fewer than five total H2 + H3 headings on a substantive page.

WordPress's block editor (Gutenberg) makes heading levels easy to set per block — but easy to set wrong. A common pattern: the page title is automatically rendered as H1 by the theme, then the editor adds another H1 inside the content because the writer chose "Heading 1" from the block dropdown. Now the page has two H1s, which AI parsers treat as a hierarchy violation.

The fix

Inside any blog post or page, the content body should start at H2. The page title (the post title) is the H1, controlled by the theme. Inside the content, use H2 for major sections, H3 for subsections, H4 only when needed. Use the Outline tool in the block editor (W key in Gutenberg) to see the actual heading structure of the page — if you see two H1s or any skipped levels, fix them.

For non-Gutenberg setups (Classic Editor, page builders like Elementor or Divi), the same rule applies but the editing surface differs. In Elementor, every Heading widget has a Tag dropdown — use H2 inside body content. In Divi, the Text module's heading tags are set via the Design tab.

How Do You Write Meta Descriptions That Pass AI Parsing?

Meta descriptions are no longer just for the Google SERP snippet. AI engines use them as a quick read on what the page is about before deciding whether to surface it in answer summaries. Two failure modes are common in WordPress: descriptions shorter than 100 characters (the audit's minimum bar for substance), and descriptions written as vague brand statements ("We are a leading provider of...") rather than concrete claims.

A meta description that passes the audit's "concrete claim" check mentions at least one of: a specific category, a named differentiator, or a geographic scope. For a Singapore aesthetic clinic, "Award-winning aesthetic treatments in Singapore" passes because it names category + geography. "A leading clinic dedicated to your beauty journey" fails because nothing is concrete enough for an AI engine to parse what the page is actually about.

How Do You Add Organization Schema in WordPress?

Organization schema tells AI engines the canonical facts about your business: legal name, URL, logo, contact info, social profiles. It's the foundation for entity recognition — being identified as a specific, real business rather than a generic web page. Without it, AI engines often fail to link your domain to your brand in their internal knowledge graphs, which means they cite competitors instead of you when answering questions in your category.

WordPress sites without an SEO plugin typically have no Organization schema at all. With Yoast or Rank Math installed, Organization schema is generated automatically — but the inputs need to be filled in correctly during the plugin's initial setup wizard, which many sites skip.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Organization",
  "name": "Your Business Name",
  "url": "https://yourdomain.com",
  "logo": "https://yourdomain.com/wp-content/uploads/logo.png",
  "sameAs": [
    "https://linkedin.com/company/your-business",
    "https://twitter.com/your-business"
  ],
  "contactPoint": {
    "@type": "ContactPoint",
    "telephone": "+65-6000-0000",
    "contactType": "Customer Service"
  }
}
</script>

How Do You Add Service or Product Schema in WordPress?

Service or Product schema is the second tier of entity recognition. Organization schema says "we exist." Service/Product schema says "we offer this specific thing." For service businesses, every major service page should have its own Service schema block. For product businesses (WooCommerce or otherwise), every product page should have Product schema with at minimum a name, description, and image — ideally with offers, brand, and aggregate rating where applicable.

WooCommerce automatically generates Product schema on product pages. Service schema is rarely automatic and almost always needs explicit setup.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Service",
  "name": "Performance Marketing",
  "provider": {
    "@type": "Organization",
    "name": "Your Business Name"
  },
  "areaServed": "Singapore",
  "description": "Senior-led performance marketing for B2B and consumer brands.",
  "serviceType": "Marketing Services"
}
</script>

Should You Still Use FAQ Schema After Google's May 2026 Deprecation?

Yes — but the rationale has shifted.

Before May 7, 2026, the case for FAQ schema was straightforward: it produced an expandable Q&A block in the Google SERP, which captured more visual real estate and increased click-through rates for ranking pages. That benefit is now gone. Google has stopped rendering FAQ rich results in search and will end Search Console and API support over the following three months.

The schema itself is still valid markup. Google has confirmed that sites can retain FAQ schema without negative indexing consequences. And there is reasonable evidence — though Google has not confirmed it — that ChatGPT, Claude, Perplexity, and other LLMs continue to parse structured Q&A content when forming their answer summaries. For a page whose primary value is a question-and-answer format, FAQ schema is still a legitimate marker of that structure for non-Google AI surfaces.

The decision rule is simpler than it used to be. If your page genuinely answers common questions in your category, mark it up with FAQPage schema. If you were adding FAQ schema as a SERP-snippet hack on pages that were not really Q&A content, stop — the SERP benefit is gone and the AI benefit is conditional on the content actually being a Q&A.

How Do You Fix Canonical Issues in WordPress?

WordPress generates canonical tags automatically, which solves the problem for most sites — but creates a different problem when you don't want the automatic canonical. Two common scenarios. First: paginated archive pages, where WordPress canonicalises every page to itself, telling AI engines that page 2 of your blog archive is a distinct canonical resource, which is rarely what you want. Second: pages syndicated from another source, where the canonical should point to the original, not your local copy.

Both are fixable per-post in any of the three major SEO plugins.

How Do You Make Your Content Concrete Enough to Be Cited?

This is the AI Readability check Google would say matters most. The audit scores three citation-quality dimensions, each judged by Claude on the actual content of the homepage and key service pages: specificity (does the page name actual products, services, or methodologies?), differentiation (does it clearly distinguish from alternatives?), and credibility (does it name real clients, results, or proof points?).

WordPress doesn't make this hard or easy — it's a content question, not a CMS question. But two patterns are common on WordPress sites and worth calling out specifically.

Generic homepage hero copy. The default of many WordPress themes is a hero block with a vague tagline ("Your trusted partner in growth") and a CTA button. AI engines reading that page have nothing concrete to extract. Replace with a specific claim: what you do, for whom, with what outcome. "Senior-led performance marketing for SEA consumer brands. Most engagements move CAC down 30%+ in 12 weeks." That sentence has category, audience, geography, and a quantified outcome — four citation hooks where the previous version had zero.

Service pages that describe the service but not the engagement. "We do SEO" is a category statement. "We work with five clients at a time, all senior-led, with monthly reporting against a single agreed KPI" is a methodology — specific, differentiated, and citable. AI engines summarise pages by extracting concrete claims. The page with no concrete claims gets summarised with the brand's name and nothing else.

Yoast's "Cornerstone content" flag and Rank Math's similar setting tell the plugin which pages matter most for SEO ranking. These pages are also where AI Readability investment pays off the fastest. Concentrate concrete-claim revisions there first.

How Do You Verify Each Fix Is Working?

Every fix above can be verified within minutes using three free tools. Treat verification as part of the fix — a schema change that didn't render the expected markup, or a robots.txt update that didn't propagate, is worse than no change because it gives false confidence.

Schema validation. Use Google's Rich Results Test (search.google.com/test/rich-results) for any page where you added schema. The tool fetches the page, runs Google's parser, and reports detected schema types and any errors. For deeper validation, the official Schema.org validator at validator.schema.org catches strict-mode errors that the Google tool sometimes hides.

Robots.txt verification. Visit https://yourdomain.com/robots.txt in an incognito window. If you added a static robots.txt file, you should see your block. If WordPress is still serving the virtual one (you see only User-agent: * and Disallow: /wp-admin/), your plugin override isn't applied or your static file isn't being served. Check file permissions and plugin save state.

End-to-end AI Readability score. Run the free Kaliber AI Readability Checker against your homepage. It parses for the specific signals AI engines consume — robots.txt AI-bot allow, OG tags, Organization schema, FAQPage schema, headings, citation quality — and surfaces what's missing in plain language. Each fix from this guide should move the score visibly. Most teams ship the top three fixes in a single sprint and see scores move 15-30 points.

What Should You Fix First on a WordPress Site?

The order that produces the largest score gain per hour of work, based on the current audit weights:

First (within 10 minutes): Update robots.txt to explicitly allow AI crawlers. This is the highest-weighted single check in the audit and the fastest fix on the list.

Second (within 30 minutes): Verify Organization schema is present and correctly populated. If using Yoast/Rank Math/AIOSEO, walk through the setup wizard or the equivalent settings panel. Add a logo, social profile URLs, and contact info.

Third (within an hour): Audit the homepage hero and top three service or product pages for concrete claims. Rewrite vague taglines and category statements to include specific differentiators, named methodologies, or quantified outcomes. This is the citation-quality work and it's the highest-impact change for getting AI engines to summarise your business correctly.

Fourth (within an hour): Fix heading hierarchy on top pages. Run the H1-count check, restructure where needed.

Fifth (within an hour): Rewrite meta descriptions on top pages to be 120-160 characters with at least one concrete claim each.

The remaining fixes — Service/Product schema on individual service pages, FAQPage schema where appropriate, canonical adjustments — are useful but smaller in score impact and can be tackled over subsequent sprints. The first five above account for the bulk of the typical score gain.

The Compounding Effect of Doing This Once

None of the fixes above require ongoing maintenance once they're correctly set up. The robots.txt block sits in your site root forever. Organization schema renders on every page via the plugin. Heading hierarchy stays fixed unless someone explicitly reverts it. Meta descriptions, once written well, persist.

That's the asymmetry that makes AI Readability work worth doing carefully on the first pass. The audit score is not a one-time vanity number. It's a structural property of the site — if you score well today, you score well in every AI search summary, every LLM citation, every voice assistant query that surfaces your category for the next several years. Sites that scored poorly in March often score poorly in October because nothing changed.

If your audit surfaced fixes that need expertise beyond a CMS settings panel — a custom theme that resists plugin schema, a page builder that won't let you control headings cleanly, or content that needs strategic rewriting rather than tactical editing — that's the work Kaliber takes on. Get in touch if you'd like a senior-led walk-through of your specific site, or read more about the GEO & AI Search service.

Frequently Asked Questions

Does WordPress need a special plugin for AI Search optimization?

No. Google's May 2026 AI Search guide explicitly states that no special markup, plugin, or file is needed to be discoverable by AI search. The same SEO fundamentals — crawler access, schema markup, headings, meta descriptions, and concrete content — that work for traditional search also work for AI Overviews, AI Mode, and other generative AI surfaces. A general-purpose SEO plugin like Yoast, Rank Math, or AIOSEO covers most of what's needed.

Should I still use FAQ schema after Google dropped FAQ rich results in May 2026?

Yes, but with realistic expectations. Google announced on May 7, 2026 that FAQ rich results no longer appear in search. The schema itself remains valid markup and won't cause indexing issues. While Google has not confirmed whether FAQ schema helps with AI Overviews or AI Mode, other large language models including ChatGPT, Claude, and Perplexity parse structured Q&A content effectively. Keep FAQ schema on pages where it adds real informational value, but stop expecting Google rich-result visibility from it.

How do I allow AI crawlers like GPTBot and ClaudeBot in my WordPress robots.txt?

Add explicit user-agent lines for each AI crawler you want to allow. WordPress's default virtual robots.txt is minimal, so most plugins (Yoast, Rank Math, AIOSEO) include a custom robots.txt editor. The blocks in this article show the lines to add. Without these lines, some AI crawlers will respect a generic User-agent: * Disallow: rule that may have been set by another plugin or theme.

Which WordPress SEO plugin is best for AI Search?

All three major plugins — Yoast SEO, Rank Math, and All In One SEO (AIOSEO) — cover the structural fundamentals that AI search rewards: meta tags, OpenGraph, canonicals, sitemaps, and basic schema. Differences are at the margin. Rank Math has the deepest free-tier schema support. Yoast has the most mature WooCommerce integration. AIOSEO is the most opinionated UX. None of them are required if you can write schema and meta tags directly in your theme, which is a valid option for developer-led sites.

How do I verify my schema is actually working after I add it in WordPress?

Use three tools in sequence. First, Google's Rich Results Test (search.google.com/test/rich-results) confirms your schema is valid and which types are detected. Second, the Schema Markup Validator at validator.schema.org gives strict parse errors that the Google tool sometimes hides. Third, run the free Kaliber AI Readability Checker against your page — it parses for the specific schema types AI engines look for and surfaces what's missing in plain language.