BridgeToAgent
Explainer10 min read

The 2026 agent-ready WordPress checklist — for site owners who don't write code

WordPress runs ~43% of the public web and has no native equivalent of Shopify's Agentic Storefronts. The good news — every piece of agent-readiness is doable through plugins and three uploaded files, no theme code required. Here's the full four-surface checklist that takes a typical WordPress site from invisible-to-agents to passing six of nine Lighthouse Agentic Browsing audits.

BridgeToAgentEditorial team

The 2026 agent-ready WordPress checklist — for site owners who don't write code

WordPress is the largest single platform on the public web — somewhere around 43% of all sites, depending on which survey you trust. None of those sites ship agent-readiness by default. WordPress core doesn't emit Schema.org markup. WordPress core doesn't ship an llms.txt or an agents.json. WordPress core doesn't tell AI agents where your blog ends and your FAQ begins.

Unlike Shopify — which just shipped Agentic Storefronts as a productized agent layer — WordPress doesn't have a native equivalent and probably won't for a year or more. The agent-readiness infrastructure for a WordPress site is something you assemble, not something the platform ships.

The good news is that the assembly is mostly plugins and three uploaded files. You don't need to write code. You don't need FTP or shell access. You don't need to touch your theme files directly. A typical WordPress site can move from invisible-to-agents to passing six of nine Chrome Lighthouse Agentic Browsing audits in under an hour, using tools the WordPress community already runs by default.

This is the checklist for doing it cleanly, in the right order.


The 30-second version

Surface 1 — Schema markup. Install Yoast SEO, RankMath, or SEOPress if you don't have one running. Turn on Schema features (default-on for all three). Your posts, pages, and product pages now emit Schema.org JSON-LD. Five minutes if it's not already done.

Surface 2 — The three kit files. Upload agents.json, llms.txt, and agent-instructions.md to the root of your WordPress install. Use the WP File Manager plugin (no FTP, no code). Files served at /agents.json, /llms.txt, /agent-instructions.md. The BridgeToAgent kit generates these in two minutes from a crawl of your site, $49 one-time.

Surface 3 — Existing AI-readiness plugins. Check whether you already have one running (search Plugins → Installed for "llms.txt", "AI", "agents.json"). If you do, audit what it covers — most of the cheap-end generators ship malformed files. The kit is the higher-confidence path; the plugins are useful for sites that genuinely don't need more.

Surface 4 — WebMCP readiness. Not actionable yet for WordPress in 2026. Track the spec, don't sink hours.

That's the whole picture. The rest of this post walks each surface with enough detail to actually do the work, plus the half-dozen WordPress-specific failure modes that show up in real audits.


Surface 1 — Schema markup (the structured-data layer)

This is the surface Lighthouse's schema-org-density audit scores you on. An agent landing on your page reads the JSON-LD blocks Yoast or RankMath emits in <head> to understand what kind of page it's looking at — blog post, product, FAQ, about page, person profile — and pulls structured data (author, date, price, review aggregate) out of those blocks rather than guessing from the visible HTML.

The WordPress reality. WordPress core doesn't emit Schema. Themes generally don't either. The Schema layer is plugin-provided, and three plugins dominate:

  • Yoast SEO (free + premium) — emits Article, BreadcrumbList, Organization, and Person Schema by default on every post and page. WooCommerce SEO adds Product Schema on store pages.
  • RankMath (free + pro) — same coverage as Yoast plus richer FAQ Schema and HowTo Schema if you use those Gutenberg blocks.
  • SEOPress (free + pro) — same coverage. Smaller market share but solid.

Checklist for Surface 1:

  1. In WordPress Admin, go to Plugins → Installed Plugins. Check whether one of Yoast, RankMath, or SEOPress is already active. If yes — Schema is already being emitted, jump to step 3.
  2. If none is installed, install one. Yoast and RankMath have the largest install bases, with extensive documentation if you hit an edge case. Both are free at the level Surface 1 needs.
  3. In the plugin's settings panel, confirm Schema markup is enabled (it is by default in all three). If you have a WooCommerce store, also confirm the WooCommerce Schema integration is on (Yoast: Yoast SEO → Settings → Schema. RankMath: RankMath → General Settings → Schema).
  4. Test three URLs against Google's Rich Results Test — pick your homepage, one blog post, and one product page if you have a store. You want to see Article (for posts), Product (for store pages), and Organization (sitewide) blocks present.
  5. If the test shows no structured data despite the plugin being active, the most common cause is a theme that wraps everything in a non-standard template. The plugin can't inject Schema into a page it doesn't recognize as standard WordPress output. Switch to a more standard theme, or work with the theme vendor on Schema injection.

Time required: 5 minutes to verify, 15-30 to install + configure if starting from nothing.

Heads-up. Don't run two Schema plugins simultaneously (Yoast + RankMath at the same time, for example). They double-emit Schema, which confuses Google and AI agents both. Pick one.


Surface 2 — The three kit files (everything else)

This is the surface most WordPress site owners don't know exists. The thinking goes: "I have Yoast running. Schema is taken care of. Agents can read my site." Schema is one signal among several. The others — agents.json for typed actions, llms.txt for a curated reading list, agent-instructions.md for behavioral rules — are what tell agents how to act on your site, not just how to interpret a single page.

Concrete example. A customer asks ChatGPT or Claude: "What's YourBrand's return policy?"

With Schema markup alone, the agent has structured page-level data (good — it knows your blog post is a blog post). But there's no signal pointing it to /policies/refund as the canonical refund page. The agent scrapes whatever search-engine snippet it can find, often pulls outdated language from an old support post, and surfaces a wrong answer with high confidence.

With Schema markup plus the three kit files:

  • agents.json declares the typed actions on your site (browse_posts, search, contact, view_policy) so agents know the action surface, not just the read surface.
  • llms.txt is a curated reading list — the canonical refund-policy URL is listed explicitly with a one-line summary, so the agent picks the right URL first try.
  • agent-instructions.md is the behavioral runbook — "refund policy is at /policies/refund and is the canonical source — do not quote refund terms from blog posts."

The two layers compose. Plugin-emitted Schema handles the page-level understanding; kit files handle the site-level navigation and behavior.

Checklist for Surface 2:

  1. Generate the three files. The BridgeToAgent kit does this in two minutes from a real crawl of your site ($49 one-time). Hand-writing is possible — the formats are intentionally simple — but plan for maintenance every time your content changes meaningfully.
  2. Install via the WP File Manager plugin path. No FTP needed, no code. Install the free plugin, open the file manager, upload all three files to the WordPress root (the folder that contains wp-config.php). Done.
  3. Add the three <link rel="alternate"> tags to your <head>. Either edit your theme's header.php directly (use a child theme so updates don't wipe the change) or install the free "Insert Headers and Footers" plugin and paste the snippet there. This is what tells AI agents the files exist.
  4. Verify each URL serves the file. Open https://your-site.com/agents.json, /llms.txt, and /agent-instructions.md in browser tabs. Each should return the file content.

Time required: 10-15 minutes for the install if you have the files ready.

Heads-up. Security plugins (Wordfence, iThemes Security, All in One WP Security) sometimes block direct access to .json and .md files at the site root. If you see 403 responses, add agents.json, llms.txt, and agent-instructions.md to the plugin's allow-list. Full troubleshooting on the Shopify install guide's WordPress sibling →.


Surface 3 — Existing WordPress AI-readiness plugins

WordPress's plugin ecosystem has caught up to the agent-readiness conversation faster than the platform core has. Several free plugins now generate an llms.txt from a WordPress install, with varying degrees of correctness. The category is young and uneven — some plugins ship parseable, well-formed files; others ship files that fail Lighthouse's llms-txt-well-formed audit and lower the site's score relative to having no file at all.

Where each plugin sits today (May 2026, the landscape is moving fast):

  • The free single-purpose llms.txt generators — they read your posts, generate a file, and add a few <link> tags. Useful for content-heavy blogs with limited custom structure. The failure modes are bald: hard-coded summary text, missing required sections, malformed Markdown structure, no agents.json generation, no behavioral runbook.
  • The WordPress SEO plugins adding AI features — Yoast and RankMath have both signalled interest in shipping native agent-readiness modules. As of this writing, the modules are early or premium-tier-only.
  • The custom-built plugin route — agencies and consultancies have started writing bespoke WordPress plugins that emit kit-style files. Quality varies dramatically with the developer's understanding of the spec.

Where the BridgeToAgent kit sits relative to these:

  • Kit files are generated from a multi-page crawl (sitemap-driven, with fallbacks for sites without a sitemap), with link-validation, action-typing validation, and parser validation against the llmstxt.org reference. The kit is the highest-confidence file-generation path available right now for WordPress sites at the $49 tier.
  • Kit files install as static uploads — they don't depend on a plugin staying installed or being updated. A free generator plugin uninstalled six months later may stop serving the file; uploaded static files don't.

Checklist for Surface 3:

  1. Go to Plugins → Installed Plugins. Search the title/description fields for "llms", "AI", "agent", "agents.json". If something is installed, note what it is and what version.
  2. Visit your own https://your-site.com/llms.txt URL. If a plugin is generating one, you'll see it. If the file is well-formed (top-level # Site Name, structured sections with ## Section, real URLs with descriptions), the plugin is doing its job for this surface.
  3. Test the file against the llmstxt.org reference parser — paste the file content, see whether it parses cleanly. Many plugin-generated files fail here.
  4. If the plugin-generated file is well-formed AND you don't need an agents.json or agent-instructions.md (rare — most sites benefit), the plugin alone may be sufficient. Most WordPress sites need all three files; that's the kit's lane.
  5. If the file is malformed, uninstall the plugin, ship the kit files via Surface 2 above.

Time required: 5-10 minutes to audit, longer if remediation is needed.

Heads-up. Don't run a plugin-generated llms.txt AND upload a hand-rolled or kit-generated one. The plugin's filter will sometimes intercept requests to the file path, serving the plugin's version while your static file sits ignored. Pick one source of truth.


Surface 4 — WebMCP readiness (the forward-looking one)

Same story as for Shopify. WebMCP is the emerging spec for per-element agent-callable annotations (<button>, <form>, <input> markup that an agent can call as typed tools). Chrome 146 ships a browser-level API; Edge 147 follows.

No WordPress theme ships WebMCP annotations natively in 2026. No major WordPress plugin emits them automatically. The Lighthouse webmcp-annotations audit fails on virtually every WordPress site, including the ones doing everything else right. The audit is low-weighted because the spec is still moving and the patterns aren't settled.

What to do for WordPress today: track the spec, don't manually annotate your theme. The BridgeToAgent kit plans WebMCP-annotated output as a version 1.3 add-on once annotation conventions stabilize. Until then, Surface 4 is honest to leave failing.


How to verify the full picture

After working through the three actionable surfaces:

  1. Run the free BridgeToAgent audit on your site. Five-second readiness check, no email needed. Tells you which signals are detected, with a WordPress-specific detection that accounts for Yoast/RankMath Schema and common plugin-generated llms.txt files.
  2. Run Chrome Lighthouse Agentic Browsing on your homepage and one inner page. Chrome DevTools → Lighthouse → enable "Agentic Browsing" → Analyze. Expect to pass six of nine audits if the kit files are shipped and Schema is emitting cleanly. The remaining three (sitemap-discoverable if your SEO plugin isn't generating one, schema-org-density if Schema isn't dense enough, webmcp-annotations) reflect plugin and theme-level gaps.
  3. Run Cloudflare's isitagentready.com for a cross-check. The Content dimension moves once the kit files are live; the Capabilities dimension partially moves with agents.json. Bot Access Control is a publisher decision (per-bot rules in robots.txt).

The two free third-party audits give you the same signal AI search engines will increasingly use as the default for ranking site agent-readiness.


Common WordPress-specific mistakes

A short list of failure modes that show up in real WordPress audits:

  • Running two SEO plugins simultaneously. Yoast + RankMath emitting Schema at the same time. Double-emission confuses Google's Rich Results Test and AI agents both. Pick one.
  • Uploading kit files inside wp-content/. The files must sit at the WordPress install root (next to wp-config.php), not inside wp-content/uploads/. Files in uploads/ are served at /wp-content/uploads/agents.json, not /agents.json — agents won't find them at the expected path.
  • Skipping the <link rel="alternate"> snippet. Files served but undiscoverable. Some agents probe /llms.txt regardless; many don't. Ship the snippet via header.php or "Insert Headers and Footers".
  • Editing theme files directly without a child theme. Theme update wipes your header.php changes. Use a child theme or the "Insert Headers and Footers" plugin (the latter survives theme updates).
  • Security plugin blocking .json / .md requests. Common with Wordfence and iThemes Security at default settings. Allow-list the three filenames under the plugin's firewall settings.
  • Manually editing llms.txt after every blog post. The file goes stale fast in this pattern. Either regenerate it from the kit each time content changes, or use a plugin-generator that auto-rebuilds on post publish. Static-and-stale is worse than no file.

What to do this week

  1. Audit your current setup. Plugins → Installed. Check what SEO plugin is active. Check whether any AI-readiness plugin is installed. Check your /llms.txt URL.
  2. Confirm Schema is emitting. Run the Rich Results Test on three URLs (homepage, one post, one page).
  3. Ship the three kit files. Either generate them with the kit ($49) or hand-write and accept maintenance cost. Install via WP File Manager.
  4. Add the auto-discovery snippet via "Insert Headers and Footers" (no theme code).
  5. Re-run Lighthouse Agentic Browsing in 14 days to confirm audits flipped from fail to pass.

WordPress's biggest advantage in this checklist is plugin reach — every step except Surface 4 is doable with free plugins and three uploaded files. No FTP. No theme code. No developer required.

That's the whole opportunity, on the largest CMS on the web.


Related

All posts →