What is
GEO?
The question every AI engine has to answer before it can answer anyone else's: who do I trust enough to name? Generative engine optimization is the discipline of being that name.
Generative engine optimization (GEO) is the practice of structuring a business's website and public data so AI engines — ChatGPT, Perplexity, Claude, Google AI Overviews — retrieve it, trust it, and cite it when answering a question a customer asks.
On this page
GEO vs. SEO
Classic SEO optimizes for a list a human scrolls. GEO optimizes for an answer a machine composes. That single difference moves nearly every lever:
| Classic SEO | GEO | |
|---|---|---|
| Unit of visibility | A ranking — position on a results page | A citation — being named inside the answer |
| The judge | A ranking algorithm ordering links for a human to choose from | A language model deciding which few sources it trusts enough to name |
| Winning outcome | The click | Being the recommendation — often with no click at all |
| Core surfaces | Titles, links, content, page speed | All of that, plus structured data depth, entity consistency, llms.txt, and answer-shaped pages |
| Failure mode | Page two — seen by almost no one | Not retrieved at all — invisible, with no page two to fall to |
The two are not rivals: an engine cannot cite what it cannot crawl, so competent technical SEO is the floor GEO stands on. But a site that stops at classic SEO is optimized for a judge that is no longer the only one in the room. The Journal covers the sharp end of this shift in the death of page two.
How engines decide who to cite
Every AI answer is the visible end of a pipeline: the engine interprets the question, retrieves candidate sources, weighs how much it trusts each, and composes an answer naming the few that survive. A business can lose at any stage — unknown to the engine, retrieved but unparseable, parsed but contradicted by inconsistent public data. GEO is the work of winning each stage on purpose: being findable, being machine-readable, and having every public fact about the business agree with every other.
The long version, with the mechanics spelled out, is in the Journal: How AI engines decide who to cite. For what this looks like done for a real business, see the Nostrand Optical case study — four valid Google rich results on launch day — and Brooklyn BJJ Lessons, cited first in ChatGPT for its core prompt within 41 days.
The terms that matter
- Citation the unit of visibility
- An AI engine naming or linking a specific business or page in its answer. In AI search, a citation is the unit of visibility — the equivalent of a first-page ranking in classic search.
- Retrieval the gate
- The step where an engine fetches candidate pages and data sources before composing an answer. A page that cannot be retrieved — blocked, unparseable, or unknown to the engine — cannot be cited, no matter how good it is.
- AI Overviews google's answer block
- Google's AI-generated answer shown above classic results, composed from sources it selects and links. One of the three surfaces a local business most needs to appear in, alongside ChatGPT and Perplexity.
- Schema.org structured data machine-readable facts
- Markup (usually JSON-LD) that states facts about a business — name, address, hours, services, reviews — in a vocabulary engines parse directly instead of inferring from prose. The Journal's seven schema types for local businesses is the practical starting list.
- llms.txt the ai-era sitemap
- A plain-text file at a site's root giving AI engines a curated, structured summary of the site: what the business is, what it offers, which URLs are canonical. The AI-era counterpart of robots.txt and sitemap.xml.
- NAP consistency entity trust
- Keeping a business's Name, Address, and Phone number identical everywhere they appear — website, Google Business Profile, directories. Inconsistent public facts make an engine less confident, and less likely to cite.
- AEO the synonym
- Answer engine optimization — a near-synonym of GEO used by part of the industry. Both describe optimizing to be the source AI answers cite. GEO is the term used in the research literature, so it's the one Signal uses.
Questions, answered
Is GEO different from SEO?
They overlap but aim at different judges. SEO optimizes to rank in a list of links a human scrolls; GEO optimizes to be retrieved and cited inside a single composed answer. Good technical SEO is a prerequisite for GEO — an engine can't cite what it can't crawl — but GEO adds structured data depth, retrievable answer-shaped content, entity consistency, and surfaces like llms.txt that classic SEO never needed.
Is GEO the same as AEO?
Effectively yes. Answer engine optimization (AEO) and generative engine optimization (GEO) both describe optimizing to be the source that AI-generated answers cite. GEO is the term used in the research literature, so it's the one Signal uses.
Can a business pay to be cited by ChatGPT or Perplexity?
No. There is no ad product that buys a citation inside an AI answer for a local business query. Citations are earned through retrievability and trust: structured data the engine can parse, consistent public facts about the business, and content that actually answers the question asked.
How long does GEO take to show results?
It varies with the engine, the competition for the prompt, and how retrievable the site was to begin with — AI engines re-crawl and re-compose answers on their own schedules, so honest practitioners give ranges, not guarantees. In Signal's published case study, Brooklyn BJJ Lessons was cited first in ChatGPT for its core prompt 41 days after launch.
Does schema markup actually matter for AI search?
Yes. Structured data states facts in a vocabulary engines parse directly instead of inferring from prose — which is precisely what a retrieval system needs before it will confidently name a business. It is the cheapest, most controllable GEO surface a business owns.
The fastest way to see GEO concretely is to watch it fail or succeed for your own business, live.
Take the free 20-minute AI audit →