Перейти к основному содержимому

SEO & GEO

Search optimization for traditional search engines and AI-powered generative engines (ChatGPT, Perplexity, Claude). Dual strategy preparing for shift in search behavior.

On this page

What is it?

Search optimization for traditional search engines and AI-powered generative engines (ChatGPT, Perplexity, Claude). Dual strategy preparing for shift in search behavior.

What is it for?

  • Traditional SEO (Google, Bing)
  • Generative Engine Optimization (AI search)
  • Meta tags and Open Graph
  • Structured data (JSON-LD)
  • Sitemaps

Why was it chosen?

Channel Current State Trend
Traditional (Google) 14B searches/day Stable
AI (ChatGPT, etc.) 66M prompts/day Growing (4.4x conversion)

Gartner predicts 50% drop in traditional traffic by 2028. Optimize for both.

Stack advantages:

Component Benefit
SSR SvelteKit default—content indexed reliably
Performance Svelte compiler—fast Core Web Vitals
Meta tags <svelte:head> + Paraglide messages—Open Graph, hreflang, canonical
Hosting Vercel edge—fast TTFB
i18n URL-based subpaths (/de/, /ru/)—SEO-friendly

GEO strategies:

Strategy Impact
Statistics with sources +22% visibility
Quotations from authorities +37% visibility
Clear structure (bullets, headings) Easier extraction
Summaries ("Key takeaways") Signals extractability

Known limitations

Traditional SEO:

  • Requires ongoing content updates
  • Backlinks still matter (hard to control)
  • Core Web Vitals requires optimization work
  • Schema markup needs manual maintenance

GEO (AI search):

  • Emerging field—strategies evolving
  • No direct control over AI training data
  • Citation tracking tools are paid/limited
  • Reference rates not guaranteed

Monitoring challenges:

Tool Purpose Cost
Profound AI visibility Paid
Semrush AI Prompt testing Paid
Manual Query AI directly Free

Implementation priority:

  1. SEO foundation (meta tags, sitemap, JSON-LD)
  2. GEO enhancement (stats, citations, structure)
  3. Active GEO (track reference rates, optimize)

Implementation

Canonical + hreflang

src/routes/[[locale=locale]]/+layout.svelte emits a self-referential canonical per locale — /de/blog canonicals to /de/blog, never to /blog. De-localizing the canonical would tell Google translated pages are duplicates and suppress them from the index.

hreflang alternates cover all locales + x-default (pointing to the en/unprefixed URL). The sitemap repeats these alternate links so both signals are consistent.

See docs/blueprint/i18n.md §Canonical and §Hreflang Tags for the pattern.

Sitemap

src/routes/sitemap.xml/+server.ts — single site-wide sitemap, dynamic (DB-backed). It cannot be prerendered because it includes published blog posts whose list changes at runtime. CDN-cached for 1h via Cache-Control: s-maxage=3600. Emits 3 <url> blocks per logical page (en unprefixed, /de, /ru prefixed), each with reciprocal xhtml:link rel="alternate" hreflang tags + x-default. <lastmod> only on blog posts (from publishedAt); omitted on static pages. <priority> and <changefreq> omitted (Google ignores them).

robots.txt

src/routes/robots.txt/+server.ts — env-aware via VERCEL_ENV:

  • Production: Allow: /, blocks /api/, /app/, /de/app/, /ru/app/; absolute Sitemap: URL.
  • Preview / development: Disallow: / — blocks indexing of preview deployments (Vercel only auto-noindexes *.vercel.app, not custom-domain previews).

The old static/robots.txt was deleted — a static file shadows the dynamic route.

Related

← Back to Stack