Structured data used to be treated as a marketing SEO checkbox — a JSON-LD block someone adds after the site is built. As AI-powered answer engines become a real discovery channel, that job is quietly becoming part of content architecture, not an afterthought.
What changed
Search is no longer only about ranking a page in a results list. Tools like ChatGPT, Perplexity, and Google's AI Overviews retrieve and synthesize content directly, often citing a specific passage rather than sending a visitor to click through. This practice is generally called Generative Engine Optimization (GEO), and it rewards content that is structured clearly enough for a retrieval system to extract facts confidently.
Why this is an architecture concern in AEM
- Content Fragment Models can encode structured facts natively — fields like author, date, credentials, or product specifications map directly to schema.org properties, so structured data generation can be automated from the model rather than hand-authored per page
- Component design affects extractability — content buried inside heavily nested, JavaScript-rendered components is harder for simpler crawlers (including many AI crawlers that do not fully render JavaScript) to extract than clean server-rendered HTML with clear headings
- EDS's server-rendered-by-default model is a natural fit for GEO — since content ships as pre-rendered HTML, it is immediately crawlable without a JavaScript execution step
Practical steps for an AEM architect
- Generate JSON-LD structured data (Person, Organization, Article, Product, FAQPage, as relevant) directly from existing Content Fragment or component data, rather than treating it as separately authored content
- Keep heading hierarchy and semantic HTML clean in component templates — this benefits both classic SEO and AI extraction equally
- Add a `robots.txt` policy that deliberately allows or blocks known AI crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended) based on a conscious content-sharing decision, not by default inheritance from a template
- Consider FAQ-style structured sections for key pages — short, self-contained question-and-answer pairs are consistently among the most reliably extracted content formats
The bigger shift
GEO is not a new discipline bolted onto SEO — it is a forcing function for doing content architecture properly: clean structure, accurate metadata, and content that says what it means without requiring a reader (human or machine) to infer context. AEM architects who already care about clean Content Fragment Models and semantic component design are, in most cases, already most of the way there.