In my previous article I explained how Generative Engine Optimization (GEO) differs from classic SEO. This one is practical: what to actually change on your website so that ChatGPT, Gemini, Perplexity and Google's AI features are more likely to use and cite your content.

None of this is a trick. Every step also makes your content better for human readers — which is exactly why it works.

The GEO checklist: answer first, structure everything, add facts and sources, show expertise, use schema markup, stay crawlable, measure and refresh

1. Answer first

AI systems extract passages, not whole pages. Make every important section quotable on its own:

  • Start each section with a direct one- or two-sentence answer, then explain.
  • Phrase headings the way people ask questions: "How much does a custom CRM cost?" rather than "Pricing Considerations".
  • Define key terms explicitly: "RAG (retrieval-augmented generation) is…".

Before: "There are many factors to consider when thinking about timelines…"

After: "A typical MVP takes 8–12 weeks. The main factors are scope, integrations and team size."

2. Structure everything

Structure helps models understand what each part of a page is about:

  • One clear H1, logical H2/H3 headings.
  • Numbered steps for processes, bullet lists for options.
  • Comparison tables for "X vs Y" questions — these are frequently reused in AI answers.
  • A short FAQ section covering the follow-up questions customers really ask.

3. Add facts, data and sources

Generic content is interchangeable; specific content is citable. The original GEO research found that adding statistics, quotations and citations made content more likely to appear in generated answers.

  • Use concrete numbers, dates and named examples.
  • Link to primary sources for claims you didn't originate.
  • Publish original data where you can: benchmarks, survey results, case-study outcomes. Unique facts are the strongest reason for a model to cite you.

4. Show real expertise

Models — and the search systems feeding them — look for signals that a real, qualified person stands behind the content:

  • A visible author name, role and bio on every article.
  • First-hand experience: "When we built X, we found Y."
  • Case studies with real context, constraints and outcomes.
  • Consistent profiles (LinkedIn, GitHub, company page) that connect the author to the topic.

5. Use structured data (schema markup)

Schema.org markup in JSON-LD tells machines exactly what a page contains. Start with:

  • Article / BlogPosting — title, author, publish and update dates.
  • Person and Organization — who you are and how to reach you.
  • FAQPage — for genuine Q&A sections.
  • Product / Service — for offerings and pricing.

Validate it with Google's Rich Results Test.

6. Stay crawlable — for the right bots

If a crawler can't read your content, nothing else matters:

  • Server-render important text. Content that only appears after heavy client-side JavaScript is harder for many crawlers to read.
  • Keep pages fast and mobile-friendly.
  • Maintain an up-to-date XML sitemap.
  • Review robots.txt deliberately. AI companies publish their crawler names (for example OpenAI's GPTBot and OAI-SearchBot, and Google's Google-Extended control). Decide which ones you allow based on whether you want visibility in their products.
  • Optionally add an llms.txt file — a proposed convention that gives LLMs a concise, Markdown overview of your site. Support across AI tools is still inconsistent, so treat it as a low-cost extra, not a guarantee.

7. Measure and refresh

You can't improve what you don't track:

  • Referral traffic: in your analytics, watch for visits from chatgpt.com, perplexity.ai, gemini.google.com and similar.
  • Prompt testing: keep a list of 20–30 questions your customers ask and check monthly which brands and pages AI assistants mention.
  • Freshness: update key pages regularly and show a visible "last updated" date — AI search favours current information.

A 30-day starter plan

Week Focus
1 List the top 20 questions customers ask; test them in 3 AI assistants and record who gets cited
2 Rewrite your 5 most important pages: answer-first sections, clear headings, an FAQ
3 Add schema markup, fix crawlability issues, review robots.txt
4 Publish one piece with original data or a detailed case study; set up AI referral tracking

Common mistakes to avoid

  • Keyword stuffing — it didn't help in the GEO research and it hurts readability.
  • Hiding content behind tabs, logins or client-only rendering.
  • Thin AI-generated filler — models are good at recognising content that adds nothing new.
  • Chasing one tool. Optimise for clarity and evidence, and you improve across all of them.

Want me to audit your site for AI search visibility, or build GEO-friendly structure into your product? Let's talk.