WordPress AEO Technical Readiness: 10 Tests Before Hiring an Agency
Eric Strate / Insights & Research
A WordPress business may have strong content and still be difficult for a search or AI retrieval system to inspect. Before paying for more articles or a fresh redesign, run ten checks that separate genuine technical barriers from assumptions about how AI assistants select sources.
Begin with the actual buyer question and URL
Choose one realistic unbranded customer question. Identify the specific page that should answer it and record its canonical URL. You cannot meaningfully test whether a page is discoverable for a question unless the page clearly describes the service, audience, and decision it supports. A website-wide score alone is not enough.
The ten-test WordPress readiness checklist
1. Inspect the public HTTP response
Check that the canonical public URL resolves successfully, normally with HTTP 200, and is not a login page, blocked error, empty shell or unexpected redirect. Repeat on desktop and mobile. Record the response URL and date; don’t assume a homepage test represents every service page.
2. Check indexing and snippet eligibility
Review the page’s indexability, canonical hints, metadata and Search Console URL Inspection where available. Google explains that an AI Overview or AI Mode supporting link must meet ordinary search indexing and snippet eligibility requirements. Indexability is necessary for those Google surfaces but not a guarantee of inclusion.
3. Review robots.txt and security controls
Confirm that the crawlers you want to permit are allowed and that hosting, CDN rules and anti-bot protections do not silently block them. OpenAI distinguishes OAI-SearchBot for search from GPTBot for training. The two controls serve different purposes. A robots.txt allowance is not proof that a firewall permits access.
4. Read the page without its styling
Inspect the returned HTML or a text-only rendering. Is the business name visible? Is the primary service stated in normal text? Are important qualifications and evidence present before JavaScript interactions? A visually impressive design can still hide critical copy behind scripts or inaccessible UI patterns.
5. Verify one coherent business identity
Compare the page’s visible business name, service coverage, location, contact details and linked profiles. Conflicting claims across header, footer, service pages and structured data create ambiguity for readers and machines. Fix the facts before tuning the markup.
6. Validate schema against the content
Inspect JSON-LD for correct syntax, type and factual alignment with the page. Use Google’s Rich Results Test for supported features and Schema.org vocabulary where appropriate. Adding FAQPage, Service, or organization properties does not create a special entitlement to AI citations. Structured data should not describe services or evidence the page does not actually provide.
7. Make supporting evidence reachable
If the page says a service works, show what can be checked: methodology, public examples, software documentation, technical notes, or client-approved outcomes. Label any internal experiment as such. Link evidence from the exact claim it supports. Unsupported promises are not improved by more headings.
8. Check internal linking and canonical destinations
Confirm the relevant service page is reachable through ordinary HTML links from useful parent pages. Inspect internal links for broken URLs, redirected historical slugs, near-duplicate landing pages and inconsistent canonical tags. The purpose is to help people and crawlers locate the correct source, not to create a page for every prompt variation.
9. Measure three different answer outcomes
Record whether an AI response names the business, cites the intended URL, or recommends the business as a choice. Those are different observations. When a platform provides neither source details nor a complete answer, report the missing evidence rather than assuming a failure.
10. Verify improvements after deployment
For each change, log the old condition, intended fix, approval, deployment date and observed new condition. Keep prompt wording, model and location as stable as feasible in follow-up tests. A citation appearing after a change is an observation, not by itself proof that the change caused the citation.
What to repair first
| Condition | Priority | Next action |
|---|---|---|
| A desired page returns a block, error or no useful content | Immediate | Restore reachability and inspect server responses. |
| The wrong URL is canonical or indexed | High | Resolve competing versions and recheck index eligibility. |
| Key business details are contradictory | High | Correct the visible facts, then align schema. |
| The page is useful but misses a real buying question | Prioritize | Add substantive, verifiable information to the relevant existing page. |
| Only a machine-specific enhancement is missing | Experimental | Test separately; do not treat it as a mandatory search requirement. |
Should WordPress sites use WARP or a special AI text file?
There is no generally required WARP, Markdown, or llms.txt format for appearing in Google AI results. WARP, the Web Answer Retrieval Protocol, is an experimental publishing structure that connects answers with supporting evidence and related questions. Its effect on retrieval or citations requires controlled testing. Treat these formats as optional research questions after the foundational checks above pass.
When does an agency make sense?
If your organization has no one who can inspect HTML, server controls, WordPress plugins, schema and analytics, an implementation partner may be useful. If you already have competent engineers, a scoped review or independent testing may be enough. the Eric Strate WordPress AEO agency page describes one managed implementation model, while its free checker provides an initial educational review. Neither substitutes for confirming actual technical access and scope.
Primary references and limitations
Google: AI features and your website; Google: Generative AI optimization guidance; OpenAI: Search crawler documentation. These public references describe eligibility and crawler controls, not a mechanism for guaranteeing model recommendations. See RankGen’s observation methodology for measurement definitions.