48% of tracked Google queries now trigger an AI Overview, up from barely 30% a year earlier. If your content isn’t cited there, that traffic goes to a competitor. Or nowhere: the user reads the AI’s answer and closes the tab.

GEO (Generative Engine Optimization) and AEO (Answer Engine Optimization) cover everything that raises your odds of being picked up in these generated answers. Plenty of agencies talk about it like some mysterious discipline. Our take: it’s 70% solid technical SEO and 30% a different way of writing. The rest is marketing vocabulary.

The uncomfortable test: does your content exist without JavaScript?

Start there, before anything else. Turn off JavaScript in your browser and reload your key pages. Whatever disappears from the screen doesn’t exist for most AI crawlers. GPTBot, PerplexityBot and ClaudeBot read raw HTML; unlike Googlebot, they don’t run your scripts.

We learned this lesson live. While auditing our own site in July 2026, we found that our blog was built client-side, in JavaScript, through the WordPress REST API. The result: to an AI crawler, the page contained nothing but a title. Zero articles, zero links. An agency selling GEO with a blog invisible to AI, that stings. That finding is exactly what triggered the site’s migration to fully static rendering.

Second nasty case spotted in the same audit: JSON-LD schemas injected at runtime by an inline script. Google eventually sees them. AI engines never do.

JSON-LD schema doesn’t make you rank. It makes you understood.

A generative engine doesn’t cite a page, it cites an entity: a brand, a person, a product, connected to each other. Your job is to make those relationships explicit in the initial HTML.

In practice, we deploy an entity graph linked by @id on every site: Organization, WebSite, Person for the author, WebPage with publication and update dates, BreadcrumbList. Then the context schemas depending on the page: Service, FAQPage, BlogPosting. All static, in the source.

FAQPage deserves a mention. Your question-and-answer sections are exactly the format answer engines prefer to slice up. Six service pages with FAQs and no schema means six missed opportunities. Few people put numbers on them, but on conversational queries (“how much does a WordPress site cost”, “how long for an ecommerce build”), the FAQ is often what gets pulled.

What AI actually cites (and what it ignores)

The March 2026 Core Update delivered numbers we had been waiting for. Pages with proprietary data or first-hand experience gained 15 to 25% in visibility. Templated content lost 30 to 50%. Generic AI content farms, 60 to 80%. Generative engines follow the same logic: they cite whatever brings information you can’t find anywhere else.

Three formats stand out in what we observe:

  • Dated, sourced data. “48% of queries as of mid-2026” gets cited. “More and more queries” doesn’t.
  • A direct answer at the top of the section. Put the answer in the first two sentences under the H2, then develop. AI slices content into blocks.
  • A named, concrete case. A specific audit, a traffic figure, a date. An AI cannot invent your experience. It’s your only truly defensible content.

On the flip side, 200-word intros that circle the topic, lists of generalities and pages with no identifiable author never show up in citations. Never.

Our GEO/AEO checklist, the one we actually use

Here’s the working list we run through on every project, in this order:

  1. Critical content visible without JavaScript (test it, don’t assume)
  2. A static, consistent JSON-LD entity graph: one @id per entity, no contradictory duplicates
  3. A real author with a Person schema, linked to the Organization and a public profile
  4. Publication and update dates in the HTML and in the schemas
  5. FAQPage on any page that contains questions and answers
  6. Direct answers under each H2, with dated, quantified data
  7. Short server response time: AI crawlers have much smaller crawl budgets than Googlebot

None of these points is exotic. Their combination is what separates a site AI engines read from a site they reproduce from memory, often badly.

FAQ

Does GEO replace classic SEO? No. Without proper Google indexing, no AI citations: generative engines lean heavily on existing search indexes. GEO adds to SEO, it doesn’t replace it.

How long before you see results? Technical fixes (static rendering, schemas) get picked up within a few weeks, at the pace of AI bot recrawls. The editorial side, as in SEO, plays out over 3 to 6 months.

How do I know if AI engines cite my site? Ask ChatGPT, Perplexity and Google AI the 10 questions your clients ask, and note who gets cited. It’s manual work but it’s the starting point. Tracking tools exist. We save them for sites that have already fixed their technical base.

Is my WordPress site a problem for GEO? Not in itself. A well-built WordPress site, with server-rendered content and clean schemas, works fine. The problem comes from the JavaScript layers some builders and plugins add, which hide content from AI crawlers.

Does your site pass the no-JavaScript test? If you’re not sure, ask for our free audit: we check how your pages actually render, your schemas and your AI citations, and we tell you what’s in the way.