Search is splitting in two. Half of it still looks like Google. The other half is ChatGPT search, Perplexity, Google AI Overviews, Bing Copilot and Claude pulling the web and answering directly. If your brand is not in the citation list, you are invisible. Generative Engine Optimization (GEO) is how you fix that.
Contents
- High-quality, citable content
- Technical foundations
- Semantic and entity optimization
- Voice and conversational queries
- Internal linking and topical authority
- AI tools for SEO analysis
- Summary
High-quality, citable content
LLMs do not rank pages — they cite chunks. A model picks 2-5 short factual passages and stitches them into an answer. Win the citation, win the traffic.
Write in short, declarative sentences. Lead with the answer, then explain. Avoid filler ("It is worth noting that…"). Long meandering paragraphs get summarised away — short factual ones get quoted verbatim.
Schema markup matters more than ever for AI. Mark up entities (Organization, Person, Article, Product) so the model can ground your brand to a node in its knowledge graph. FAQ schema is a citation magnet — Q&A blocks are exactly the format LLMs want.
Technical foundations
The technical bar has not moved much. AI crawlers still need crawlable, fast HTML.
- Speed — LCP < 2.5s, INP < 200ms, CLS < 0.1. PageSpeed Insights and WebPageTest are your meters.
- Mobile-first — most crawlers (including AI) render mobile first.
- HTTPS — non-negotiable.
- Clean URLs —
/seo/seo-for-ai-and-llms/beats/p?id=4421. - Semantic HTML — proper
<h1>,<h2>,<article>,<nav>. LLMs use the DOM structure to extract meaning.
New thing: control AI crawlers via robots.txt. The relevant user agents:
User-agent: GPTBot User-agent: OAI-SearchBot User-agent: ClaudeBot User-agent: PerplexityBot User-agent: Google-Extended
Allow them if you want citations. Block them if you do not want your content used as training data. Most B2B brands should allow — citations beat exclusion.
There is also the llms.txt proposal (answer.ai) — a curated entry point for LLMs, similar to sitemap.xml but optimized for model context. Adoption is early but cheap to ship: a single Markdown file at /llms.txt listing your most important pages with one-line descriptions.
Semantic and entity optimization
AI search runs on entities, not keywords. "Kaminski.link" is an entity. "web design agency Poland" is a query. The model needs to connect the two.
How to feed entities cleanly:
- Wikipedia presence — even a stub article massively boosts entity recognition. If you cannot get a Wikipedia page, get cited on Wikipedia.
- Consistent NAP — Name, Address, Phone identical across the web (your site, Google Business Profile, LinkedIn, industry directories).
- Schema.org Organization + Person with
sameAslinks to LinkedIn, X, GitHub, Crunchbase. This is how the model ties profiles to one entity. - Brand mentions without links count for AI even more than backlinks. Get talked about on Reddit, Hacker News, Substacks in your niche.
Long-tail queries dominate AI search because users ask full questions. Optimize for "How do I migrate WooCommerce to Shopify without losing SEO?" not "shopify migration".
Voice and conversational queries
Voice and chat are converging. Both reward natural-language Q&A formatting.
- Headings as questions: "How does Generative Engine Optimization work?"
- Answers in 1-2 sentences right under the heading, then expand below.
- Local SEO — AI Overviews lean heavily on Google Business Profile and local citations. Claim and verify it.
Internal linking and topical authority
Topical clusters work in AI search even harder than classic SEO. Pick a topic, write the pillar page, link 5-10 supporting articles to it. The model treats your site as the authority on that topic.
Link relevant pages to each other with descriptive anchor text — "see how we cover Core Web Vitals" beats "click here". External backlinks still matter, but brand mentions (with or without a link) increasingly count for entity authority. Get cited in industry publications, podcasts and Substacks.
Audit your link profile with Google Search Console, Ahrefs or Moz. Disavow obvious link-farm garbage — AI models inherit Google's spam signals.
AI tools for SEO analysis
A handful of tools earn their cost for GEO work:
Surfer SEO
Content optimization scored against ranking competitors. Tells you which entities and headings are missing. Useful, not magic — do not let it dictate every word.
Clearscope
Semantic content analysis. Surfaces related entities and terms that strengthen topical relevance. Good for editorial QA.
ChatGPT and Perplexity
Use them as the audit tool. Ask "What are the top 5 sources for [your niche query]?" and see who gets cited. If your brand is not there, you have your gap analysis.
Profound, Otterly, Peec.ai
New class of tools tracking your brand's citation share across LLMs. Worth watching but early-stage.
Summary
GEO is not a replacement for classic SEO — it is an extension. Win citations by writing short, factual, schema-marked content. Build entity authority through Wikipedia, brand mentions and consistent NAP. Allow the right crawlers. Track your citation share, not just rankings.
The window where GEO is cheap and underserved is closing fast. Brands that move now will own the citation slots when AI search becomes default.
Want your brand cited by AI search? Get my free SEO audit — I'll show you what's missing for AI visibility.
