Technical SEO for AI search

AI search does not remove the classic technical SEO prerequisites — it raises the bar on them. Answer engines retrieve from crawl snapshots. A page that is not discoverable, not crawlable, or not correctly indexed cannot be cited regardless of its content quality or schema depth. The technical layer is the floor everything else stands on.

The minimum technical pass

These are the conditions a page must meet before any content or citation work is meaningful. Failing any one of them will suppress the page in both traditional and AI search.

Status and indexability

Every primary page returns HTTP 200 — no soft 404s (200 status with “not found” body)
No accidental noindex in the meta robots tag or HTTP header
No blocking by robots.txt on content pages (wildcard rules are a common source of accidental blocks)
HTTPS with a valid, unexpired certificate and no mixed-content warnings

Canonical hygiene

One canonical URL declared consistently in: <link rel="canonical">, the HTTP header, and the sitemap
No conflicting canonicals across paginated or faceted variants
No redirect chains longer than two hops pointing to canonicals
Trailing-slash consistency: choose one form and enforce it with 301s

Sitemap accuracy

Submitted and verified in both Google Search Console and Bing Webmaster Tools
Only canonical, indexable 200 URLs are included — no noindex pages in the sitemap
Updated automatically after every publish; lastmod reflects actual content change dates
Size under the 50,000 URL and 50 MB limits; split into sub-sitemaps if needed

Structured data alignment

JSON-LD blocks validate without errors in Google's Rich Results Test
Schema type matches the actual content: service pages use Service / OfferCatalog, not Product
Every claim in structured data is also present in the visible page text
No fabricated or aspirational data in structured markup

Core Web Vitals (field thresholds)

LCP (Largest Contentful Paint) ≤ 2.5 s at the 75th percentile
CLS (Cumulative Layout Shift) ≤ 0.1 at the 75th percentile
INP (Interaction to Next Paint) ≤ 200 ms at the 75th percentile

Mobile parity

Mobile layout returns the same core content as desktop — no hidden text, no collapsed critical copy
Viewport meta tag present; touch targets ≥ 44 × 44 px

High-risk patterns for AI citation specifically

JavaScript-only rendering

If the answer lives only in JavaScript that runs after page load, most crawlers never see it. The page exists in the visual DOM but not in the HTML response body — which is what gets stored in the crawl snapshot. Server-rendering the content (SSR or SSG) is the reliable fix. Client-side hydration on top of server-rendered HTML is fine; client-only rendering with no server HTML is the problem.

Infinite scroll and interaction-gated content

Content that loads only after a scroll event, a click, or a tab selection is structurally invisible to a crawler following a single HTTP GET. Break that content into separate indexable URLs or surface it in the initial HTML response.

Framework routing that breaks canonical

Client-side navigation in SPAs can change the displayed URL without issuing a server redirect. Crawlers follow HTTP redirects, not JS pushState — so two paths pointing at the same content without a server-level canonical or redirect create a duplicate-content split.

Thin or AI-generated duplicate content at scale

Pages generated programmatically with low variation — especially pages written to target AI assistants rather than humans — are treated as low-quality by search engines and increase the risk of domain-level quality suppression. The standard is the same one that has always applied: useful, accurate, distinctly informative pages for real queries.

Semantic HTML: the overlap with accessibility

The same semantic structure that helps assistive technology also helps crawlers. A page built with proper landmarks, ordered headings, labelled images, and real interactive elements (buttons, links) is interpretable by any automated reader — human, crawler, or AI agent. This is why accessibility and AI visibility are measured in the same audit pass.

Use HTML5 landmarks: header, main, article, section, nav, footer
One h1 per page; H2 → H3 → H4 hierarchy in strict order, no skipped levels
All images that communicate meaning have descriptive alt text
Interactive elements use real <button> or <a href> — not styled divs with click handlers
Form controls have associated <label> elements

A note on structured data

Structured data is parse hygiene, not a citation lever. It helps crawlers understand what a page is. It is a good technical practice and we add it to every client site. But structured data on a page with poor rank, no third-party mentions, and JS-only rendering will not produce citations. Fix the floor before decorating the ceiling.

Get a citation gap scan

What this does not guarantee

Schema and structured data are hygiene — not citation levers. Adding structured data to a page does not cause AI engines to cite it.
No specific citation count, rank position, or AI response outcome is promised by any action described on this page.
The refund covers the delivered work, never a citation number.