GEO guide

Get cited in AI answers — on purpose.

Buyers now ask ChatGPT, Gemini and Claude what to buy. The answer names a few tools — and if yours isn’t one, there’s no page two to scroll to. This is how that layer works, and how you win it.

What it is

GEO is optimizing to be the brand AI recommends.

Generative & Answer Engine Optimization is the AI-search cousin of SEO. SEO competes for a rank in a list of links. GEO competes to be quoted inside the single synthesized answer the engine hands the buyer — citations and all.

Why it’s different

AI search isn’t blue links.

The mechanics that made SEO work don’t map cleanly onto an engine that reads, reranks and rewrites the web into one answer.

One answer

Not ten links — one verdict.

There’s no scrolling to result #7. The engine names a handful of options and moves on. You’re either in the answer or you don’t exist.

Third-party

Your site isn’t the source.

Roughly 90–95% of citations point to third-party pages — reviews, forums, press. Winning means getting into the sources the engine already trusts, not just polishing your homepage.

Fan-out

One prompt becomes many.

Before answering, the engine rewrites your question into ~8–9 sub-queries and pulls sources for each. The real surface to win is those sub-queries, not the headline prompt.

The two levers

Two ways to get cited.

Every citation clears two gates. Confusing them is the most common — and most expensive — GEO mistake.

Gate 1

Discoverability

Is your page even in the retrieval pool? This is driven by off-site authority and search-index coverage — the same forces that make a source show up at all. No amount of on-page polish helps if you’re never retrieved.

Gate 2

Selectability

Once retrieved, will your page survive the rerank and get quoted? ChatGPT cites only ~15% of the pages it retrieves. Selectability is won on-page: clear passages, fact density, schema, freshness, speed.

The gap

The source gap

Between the two sits the practical question: which third-party domains shape your category’s answers, and where are you absent? That gap is the shortest path to more citations.

What the data says

The research is in — and it isn’t old SEO.

A few independent studies now measure how AI engines actually pick what to cite. The numbers point the same way.

76% → 38%

share of Google AI Overview citations coming from the top-10 organic results, year over year. Ranking #1 no longer means getting cited.

~85%

of the pages an engine retrieves are never cited. Being in the pool is necessary, not sufficient — selection is the bottleneck.

~40%

visibility lift from the top studied on-page moves — adding quotes, statistics and cited sources — over baseline.

~25%

of U.S. ChatGPT citations come from Wikipedia and Reddit alone. A few trusted sources shape most answers.

†Sources: Ahrefs (~4M AI-Overview URLs; AI-search study); Princeton/Georgia Tech GEO benchmark, KDD 2024; 5W Research. These are associations, not proven cause — and AI citations swing month to month, so we track every metric over time with a confidence range, never as a single point.

In practice

How to actually win it.

Measure honestly

Track share of voice, mentions, position and sentiment across all three engines — multi-run, with a confidence range, because AI citations swing 40–60% month to month.

Find the source gap

Identify the third-party domains the engines cite for your category, then see exactly which ones already mention your competitors but not you.

Make pages quotable

Add 40–60 word answer nuggets and fresh dates so a model can lift your point verbatim — and earn your way into the sources that feed the answer. The studied moves that move the needle: adding quotes, statistics and cited sources lifted visibility ~40% over baseline (Princeton GEO benchmark, KDD 2024).

Go deeper.

The research behind this guide, the report that turns it into a plan, and the rest of the resource library.

Research

See where you stand in AI search — in 30 seconds.

See how AI sees you Explore the resources