"GEO" is the new acronym this year. Every agency is selling it as the next must-buy retainer. Most of what they sell is technical SEO with a coat of fresh paint.
The honest version: large-model answer surfaces — ChatGPT, Claude, Gemini, Perplexity — index the web mostly the same way Google does. They want clean HTML, real headings, structured data, and content that survives summarization. If you do SEO properly in 2026, you do GEO too.
What's actually different
Three things matter more for retrieval-by-LLM than for traditional search ranking:
- Question-shaped headings outperform statement headings. "How does this work?" beats "Our process."
- Short, declarative answers — ideally under 80 words per answer block — are quoted verbatim.
- JSON-LD schema tells the retriever exactly what type of content it's looking at —
FAQPage,HowTo,Person,Article.
The four schema tweaks
Add these four blocks to your <head>. They take an afternoon and they compound across every page you publish.
1. Person schema — on your /about page. Tells answer engines who you are.
{
"@context": "https://schema.org",
"@type": "Person",
"name": "Andrew Yong",
"jobTitle": "Web Designer",
"url": "https://andrewyong.dev"
}2. FAQPage schema — on every page with a Q&A block. Makes individual questions and answers quotable.
3. BreadcrumbList — helps the retriever understand your site's information architecture.
4. WebSite with SearchAction — signals that you have a sitemap and a search endpoint.
How to test if it's working
Ask Perplexity to summarize your home page. If it can return a coherent two-paragraph summary that names your business, what you do, and your headline value-prop, your schema and content structure are correct. If it can't, you have work to do.
What I do for clients
This work is built into every site I make, and I also run it as a standalone SEO and answer-engine retainer. See what I do if you want it done for you.
Written 2026 · 04 · 18 · all writing →
Related reading, same shelf.