The SEO + GEO playbook: rank in Google and get cited by AI

Ranking in 2026 is two jobs, not one: earn a top spot in Google and get quoted by AI assistants. The good news is they share one foundation — fast, crawlable, structured, evidence-rich pages — so you build it once and win on both surfaces. Below is the exact playbook we follow, step by step, with the real scores we hit on our own site, ashvara.io, as the worked example.

Why this is two games now

Search is splitting into two surfaces. Classic search still sends the majority of traffic, and it still rewards speed and structure. But a second surface — answers generated by ChatGPT, Perplexity, Claude, and Google's AI overviews — now handles a fast-growing share of informational queries (by various 2026 industry estimates, already a double-digit percentage of English-language informational search). And critically, the pages those engines cite increasingly differ from Google's top blue links — industry analyses in 2026 report the overlap has fallen sharply. That means you can rank #1 and still be invisible in the AI answer, or vice versa.

So you optimize for both. SEO gets you ranked; GEO — Generative Engine Optimization — gets you quoted. Here's how to do each, in order.

Part 1 — The technical SEO foundation

This is the part that's pure engineering, and it's where most sites quietly lose. Work through it in this order:

1. Pass Core Web Vitals. Google measures real-user experience on three metrics with documented "good" thresholds: LCP under 2.5s (loading), INP under 200ms (responsiveness — this replaced FID as a Core Web Vital in March 2024), and CLS under 0.1 (visual stability), all judged at the 75th percentile of real visits. The reliable way to hit them is architectural: render static HTML, optimize and size every image, and reserve space for media so nothing jumps.
2. Make the site trivially crawlable. A clean sitemap.xml with real lastModified dates, a robots.txt that points to it, and a canonical URL on every page so duplicate paths don't split your ranking signals.
3. Add structured data (JSON-LD). This is how a search engine understands what a page is, not just what it says — and it's what unlocks rich results. Mark up your organization, products, articles, breadcrumbs, and FAQs.
4. Get metadata and social cards right. A unique, keyword-bearing <title> and a description under ~160 characters per page (that's your search snippet), plus Open Graph / Twitter images so shared links look intentional.
5. Use semantic HTML and real internal links. Headings in order, meaningful link text, and contextual links between related pages so crawlers — and readers — can move through the site.

None of this is exotic. The discipline is doing all of it, on every page, by default.

Part 2 — The GEO content layer

Technical SEO makes a page eligible. GEO makes it quotable. The foundational research here — the GEO: Generative Engine Optimization study (Princeton & Georgia Tech, KDD 2024) — tested what actually makes AI engines cite a page, and the findings are concrete:

• Adding statistics lifted citation visibility ~32%, quotations ~41%, and citations to authoritative sources ~30%.
• About 44% of an engine's citations come from the first third of a page.

An AI engine doesn't reward the cleverest prose. It rewards the page that hands it a clean, sourced, quotable fact — early.

So the content rules follow directly: lead with a direct, quotable answer in the first paragraph; back claims with real numbers and named, linked sources; use clean lists and headings an engine can lift without ambiguity; and keep pages fresh, since recently updated content shows up disproportionately in AI answers. Add an llms.txt to spell out your site for language models. We go deeper on this in how to show up in AI search.

The worked example: how ashvara.io scores

We built this site to practise exactly what's above. Here's an unedited Lighthouse run on the live home page:

Lighthouse on the deployed ashvara.io home page (desktop). Accessibility, Best Practices, and SEO are a stable 100; Performance is the one metric that moves run-to-run.

The full picture, measured on the deployed, edge-cached pages:

• Accessibility, Best Practices, and SEO all score 100, with Performance in the mid-90s (96 on desktop, 91 on throttled mobile in our latest runs) — and real-world LCP ≈ 2.4s and CLS of 0 on app pages. Performance is the metric that fluctuates per run; the other three hold at 100.
• 100% static generation on Next.js, deployed to Vercel's edge — which is why the vitals are good, not a coincidence. Static HTML and aggressive image optimization remove the usual sources of slow LCP and layout shift.
• Full JSON-LD coverage: Organization (with sameAs to our App Store developer page), SoftwareApplication for every app, Service, Article, BreadcrumbList, and FAQPage.
• Crawl plumbing done right: a generated sitemap.xml with per-page lastModified, a robots.txt, canonical URLs everywhere, dynamic Open Graph images via the framework's image API, and an llms.txt for AI engines.
• GEO baked into the content: every blog post here front-loads a bolded answer, cites real studies with links, and uses lists an engine can lift — including the one you're reading.

The point isn't the score for its own sake. It's that all of it was built in from the first commit, not bolted on before launch.

Our opinion

SEO and GEO are not a marketing task you run after the site is built — they're a property of how it's built. A page that renders as fast static HTML, declares what it is in JSON-LD, and leads with a sourced answer is simultaneously a great Google result and a great AI citation. Try to add those qualities later and you're refactoring; build them in and they're free.

One caution: ignore the rumor mill. In 2026 you'll see confident blog claims like "Google now requires LCP under 2.0s." The documented threshold is still 2.5s. Build on primary sources — Google's own docs and peer-reviewed research — not on the SEO telephone game. The fundamentals move slowly; the panic cycles don't.

How Ashvara helps

We build every site this way by default — fast static rendering, complete structured data, and content engineered to be both ranked and quoted. We did it on our own site, and we've shipped 18 apps whose pages funnel from it. If you want a site that earns both the Google spot and the AI citation, that's our web development practice — tell us what you're building and we'll give you a senior read on where you stand today.

---

Sources: Core Web Vitals (web.dev / Google) for LCP, INP, and CLS thresholds; GEO: Generative Engine Optimization (Aggarwal et al., KDD 2024) for AI-citation findings. Lighthouse scores are from our own deployed pages and will vary with network conditions.