Workflow

Step 1: Keyword mapping

Map each target page to a primary keyword, 2-3 secondary keywords, and a user-intent bucket (informational, navigational, transactional, commercial).

Step 2: Implement generateMetadata in Next.js

// app/blog/[slug]/page.tsx
import type { Metadata } from "next";
import { getPost } from "@/lib/posts";

type Props = { params: Promise<{ slug: string }> };

export async function generateMetadata({ params }: Props): Promise<Metadata> {
  const { slug } = await params;
  const post = await getPost(slug);
  if (!post) return {};

  return {
    title: `${post.title} | YourBrand`,
    description: post.excerpt.slice(0, 155),
    alternates: { canonical: `https://yourbrand.com/blog/${slug}` },
    openGraph: {
      title: post.title,
      description: post.excerpt.slice(0, 155),
      url: `https://yourbrand.com/blog/${slug}`,
      type: "article",
      images: [{ url: post.ogImage, width: 1200, height: 630 }],
    },
  };
}

Step 3: Entity gap analysis

1. Collect the top 10 SERP results for your primary keyword.
2. Extract named entities from each using an NLP tool or manual review.
3. Build a union set of entities and mark which ones your content covers.
4. Fill gaps with factual, sourced statements that naturally include missing entities.

Step 4: JSON-LD structured data

Add application/ld+json script tags for the relevant schema type. See the schema-markup skill for full implementation patterns.

// components/article-jsonld.tsx
export function ArticleJsonLd({ post }: { post: Post }) {
  const schema = {
    "@context": "https://schema.org",
    "@type": "Article",
    headline: post.title,
    author: { "@type": "Person", name: post.author },
    datePublished: post.publishedAt,
    dateModified: post.updatedAt,
    image: post.ogImage,
    publisher: {
      "@type": "Organization",
      name: "YourBrand",
      logo: { "@type": "ImageObject", url: "https://yourbrand.com/logo.png" },
    },
  };
  return (
    <script
      type="application/ld+json"
      dangerouslySetInnerHTML={{ __html: JSON.stringify(schema) }}
    />
  );
}

Step 5: Set up llms.txt

Create public/llms.txt listing pages AI crawlers should prioritize:

# YourBrand llms.txt
# See https://llmstxt.org for spec

> YourBrand builds developer tools for X.

## Docs
- [Getting Started](https://yourbrand.com/docs/start)
- [API Reference](https://yourbrand.com/docs/api)

## Blog
- [How We Built X](https://yourbrand.com/blog/how-we-built-x)

For a verbose version, also ship public/llms-full.txt with inline content.

Step 6: Platform-specific GEO signals

Examples

Example 1: Blog post metadata in Next.js App Router

export async function generateMetadata({ params }: Props): Promise<Metadata> {
  const { slug } = await params;
  const post = await getPost(slug);
  return {
    title: post.title,
    description: post.excerpt.slice(0, 155),
    alternates: { canonical: `https://example.com/blog/${slug}` },
    openGraph: {
      title: post.title,
      description: post.excerpt,
      type: "article",
    },
  };
}

Example 2: Entity density improvement

Before: "Our tool helps with SEO." After: "Our tool analyzes Google Search Console data, identifies keyword cannibalization, and generates schema markup compliant with Schema.org Article and FAQ types."

Decision tree

• Page needs to rank in Google → apply full keyword placement workflow
• Page targets AI search only → focus on entity density + llms.txt + direct answers
• Page targets both → apply both workflows; JSON-LD and llms.txt are complementary
• Page is behind auth → skip SEO; focus on UX and internal search
• Content is thin (< 300 words) → expand with entity-rich paragraphs before optimizing

Edge cases and gotchas

1. Keyword stuffing penalties — Google penalizes unnatural repetition. Keep primary keyword density between 1-2%. Use semantic variants instead of exact repeats.
2. Duplicate metadata across pages — every page needs unique title and description. Use generateMetadata dynamically, never hardcode the same string.
3. JavaScript-rendered content — Google renders JS but AI crawlers may not. Ensure SSR or SSG for all indexable content. Test with curl to verify.
4. Canonical conflicts — a page with both a self-referencing canonical and a redirect will confuse crawlers. Pick one source of truth.
5. llms.txt freshness — AI crawlers cache aggressively. Update llms.txt and llms-full.txt whenever you publish or remove important pages.
6. Over-optimizing for one AI platform — each AI search engine weights signals differently. Optimize for the union of signals, not a single platform.

Evaluation criteria

• Every indexable page has a unique title, meta description, and canonical URL
• Primary keyword appears in title, H1, first 100 words, and at least one subheading
• JSON-LD is valid (test with Google Rich Results Test)
• Entity coverage matches or exceeds top 3 SERP competitors
• llms.txt exists at the site root and is current
• Page loads with SSR/SSG — content visible without JS execution
• Core Web Vitals pass (LCP < 2.5s, CLS < 0.1, INP < 200ms)
• No duplicate canonical URLs across the site

Research-backed changes

The biggest delta is Search Central. Fold the concrete changes into the operating notes, then discard the fluff.

Fresh signals

• Search Central · Google Search Central
• Google Search Central Blog · Google Search Central
• 排名更新版本記錄 · Google Search Central
• 網頁體驗 · Google Search Central

Next edits

• Add one new tactic based on Search Central.
• Rewrite one stale section with today's source language.
• Turn the top signal into a reusable agent prompt.