Skip to content

Issue 14EssayNo. 011

SEO & GEO

GEO is the new SEO: how to get cited by AI answers.

Search is splitting in two. Here's the playbook we use to make client sites the source that AI Overviews and chat assistants quote — without giving up the rankings you already have.

Maya Okafor

Head of SEO & GEO

Published
Updated
Read time
12 minutes
A strategist at a desk at night reviewing a dashboard of search and AI citation data on a large monitor.
FIG. 1 — A strategist reviewing how AI assistants cite a client's pages.Photo: Performative Web Studio
In this article
  1. What actually changed
  2. How AI answers choose sources
  3. The GEO playbook, step by step
  4. Mark up your expertise
  5. Measure what you can
  6. Key takeaways

In March, a client's organic sessions fell 18% — and their enquiries went up. Buyers had stopped clicking through. They were asking AI assistants instead, and the assistants were quoting our client.

That's the shift in one paragraph. Google's AI Overviews, ChatGPT search, Perplexity and Copilot now answer a growing share of questions before anyone sees a link. Ranking still matters, but it's no longer the finish line: the new goal is to be the source the answer is built from.

We call the work that gets you there GEO — generative engine optimisation. It isn't a replacement for SEO. It's what SEO looks like when the first reader is a model and the second is a person.

What actually changed

Classic search ranks pages. Generative search ranks passages. A model retrieves a handful of candidate documents, pulls out the paragraphs that answer the question most directly, and stitches them into a response — usually with two to five citations.

Across 40 client sites we tracked this year, the pages that earned citations shared three traits: a direct answer near the top, a named and credible author, and structured data that made the page's purpose unambiguous.

“You're no longer competing for ten blue links. You're competing to be the sentence the answer is built from.”

— Maya Okafor, Head of SEO & GEO

How AI answers choose sources

Models favour content that is easy to extract and easy to trust. That means short, self-contained paragraphs; specific numbers with dates; and signals of expertise the model has already seen corroborated elsewhere — reviews, directory listings, press mentions.

A printed article on a dark desk with several passages highlighted in lavender, beside a pen.
Fig. 2 — Cited passages are short, specific and self-contained. We audit client pages for them line by line before touching anything else.

The GEO playbook, step by step

This is the sequence we run on every SEO & GEO engagement. Steps one to four take about six weeks on a typical 50-page site; step five never really ends.

  1. Answer the question in the first 60 words.

    Lead every key page with a direct, quotable answer, then expand with detail and evidence.

  2. Make every claim checkable.

    Cite sources, date your numbers and name authors with real, verifiable credentials.

  3. Structure for extraction.

    Descriptive H2s, short paragraphs, tables and FAQ blocks that a model can lift cleanly.

  4. Mark it up.

    Article, FAQPage, Organization and Product schema in JSON-LD — consistent across the site.

  5. Earn mentions off-site.

    Reviews, directories and press coverage in the places models already trust.

Field note · Check this first

Don't block AI crawlers by accident. Check robots.txt for GPTBot, ClaudeBot and PerplexityBot before anything else — we find at least one blocked on most audits, usually by a security plugin nobody remembers installing.

Mark up your expertise

Structured data won't get you cited on its own, but it removes guesswork. At minimum, every article should carry Article schema with a real author, and your business should be described once — consistently — with Organization schema.

JSON-LD · article-schema.html
<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "GEO is the new SEO",
  "author": {
    "@type": "Person",
    "name": "Maya Okafor",
    "jobTitle": "Head of SEO & GEO"
  },
  "datePublished": "2026-10-14",
  "publisher": { "@id": "https://performativeweb.com/#org" }
}
</script>

Measure what you can

Attribution from AI answers is still patchy. We track three proxies every month: citation share across a fixed set of 50 buyer prompts, branded search volume, and referral sessions from chatgpt.com, perplexity.ai and copilot.microsoft.com in GA4.

None of them is perfect on its own. Together they're enough to see whether the work is compounding — and to stop doing what isn't.

Key takeaways

If you only remember three things

  1. Generative search ranks passages, not pages — write paragraphs that can stand alone.

  2. Credibility is a ranking factor for models: named authors, dated facts, off-site mentions.

  3. Keep classic SEO healthy. Most AI answers still retrieve from pages that already rank.

  • GEO
  • AI search
  • Schema
  • Content

Share

Portrait of Maya Okafor

Written by

Maya Okafor

Head of SEO & GEO · 11 years in search

Maya leads our search practice, from technical audits to AI-answer visibility. Before Performative Web she ran SEO for two marketplaces through four core updates. She writes here about what's working now — with the client numbers to back it.

The monthly build

Get the next essay
in your inbox.

One email a month · 4,200 readers · No spam

Rather we ran this playbook for you?

Book a GEO audit