1. Render content in HTML, not JavaScript
Vercel and MERJ's 2024 research showed that pages requiring client-side JavaScript to render their main content are systematically excluded from many AI crawlers. OpenAI's GPTBot and Anthropic's ClaudeBot don't execute JavaScript the way Googlebot does.
Fix: use static site generation (Next.js output: 'export'), server-side rendering, or traditional HTML. The content must be in the initial HTML response, not assembled in the browser.
2-5. Four content patterns that move citations
These four moves compound — none of them is a magic bullet, but together they shift a page from "crawled" to "cited."
- Lead each page with a 50 to 100 word answer paragraph. Plain prose, no marketing fluff, names the entity and answers the implied question.
- Use H2 headings phrased as questions or direct claims ("What does AI automation cost?" beats "Our Pricing").
- Cite verifiable statistics inline with the source name ("Princeton 2024 GEO paper", "Local Falcon May 2025") — never "studies show."
- Add an FAQ block of 5+ Q&A pairs that match real user query language. Use FAQPage JSON-LD to mark it explicitly.
6-8. Schema, sitemap, and llms.txt
AI engines reward machine-readability. Three structural elements every site needs:
- Schema.org JSON-LD on every meaningful page. Article or BlogPosting for blog posts; Service for service pages; Organization once at the site level; FAQPage for FAQ blocks.
- An XML sitemap at /sitemap.xml listing every important URL. Updated automatically on every build.
- A llms.txt at the root curating the 20-60 highest-value URLs with one-sentence annotations. Hint, not enforcement, but signals technical competence.
9-10. Name real things
AI engines build citation graphs around real-world entities. Pages that name disambiguated entities (people, places, organizations, products) get cited more often than pages that talk in abstractions.
- Name the founder, the town, the products, the clients (with permission). "Tyler Preisser in Hays, Kansas, built a custom CRM for Astrus Insurance" carries more entity weight than "our team helped a Kansas insurance agency."
- Link out to authoritative external entities where relevant (Princeton, Local Falcon, Gartner research). Outbound citations to recognized entities improve your own citation worthiness.
11-12. Dates and identity
AI engines prefer fresh, identity-anchored content over anonymous evergreen copy.
- Add datePublished and dateModified to every Article/BlogPosting page. Update dateModified when content meaningfully changes.
- Establish a canonical author identity. Person schema for the author, sameAs links to verified profiles (LinkedIn, GitHub, professional associations). E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) signals carry weight.
Things that look right but hurt you
A few moves that look like AEO optimization but actually downweight your citations:
- Stuffing FAQ blocks with questions nobody asks. AI engines detect generic LLM-generated Q&A and penalize it.
- Schema markup that contradicts the visible content. Google penalizes structured-data mismatches.
- Hidden text or links intended to influence crawlers but not visible to users.
- Adding 50+ outbound links to appear well-researched. Citation graphs notice link-quality patterns.
- Publishing AI-generated content with no editorial pass. The phrasing patterns are detectable and discounted.
