Technical Foundation for AI Search: Structure, Speed, and Signals

Part of our AI Search series. Previous: Answer-First Content.

Content and PR tactics fail on slow, uncrawlable, unstructured sites. Most AI products retrieve from indexes built on pages Google could read and trust.

If you're not in Google's index, you don't exist in the AI answer. The fundamentals still matter — maybe even more so now.Emma Young, WordCamp 2026

When organic rankings fall, AI citations often fall in parallel — same sites, same timeline. Chasing 'GEO' on a weak technical base is decorating a house with foundation cracks.

• Resolve crawl errors, orphan pages, and accidental noindex.
• Canonical URLs and consistent internal linking — items on our post-backend SEO checklist.
• Server response times and renderability — bots and users both bounce on slow pages.

JSON-LD helps machines understand entity type, FAQs, articles, and products without guessing from layout alone.

On WordPress: use established schema plugins or theme support — validate in Rich Results Test. On Next.js: JSON-LD in layout or page components, typed where possible. Organization + WebSite on root; FAQPage on support content; Article on blog posts.

An llms.txt file at your site root tells LLM crawlers where authoritative content lives — a lightweight map, not a replacement for good HTML.

# llms.txt
> Archiom — product engineering

/docs: Technical guides
/services: What we build
/blog: Engineering articles

Pair with semantic HTML: real heading hierarchy, descriptive link text, tables for comparative data. Avoid rendering critical copy only in client-only JavaScript without SSR.

WordPress clients often have plugin bloat and duplicate schema — audit before adding more. Headless Next.js clients get control but must not skip sitemap.ts, robots.ts, metadataBase, and JSON-LD — gaps we track in docs/post-backend.md.

Next: The Monday AI Visibility Audit. Pillar.