GEO: Why Your Business Isn't in ChatGPT's Answers
When a customer asks an AI for a recommendation, something decides which businesses get named. It isn't an ad auction, and it isn't luck. Here's how Generative Engine Optimization actually works.
Ask Google "who's the best caterer near me" and you get ten blue links, a map, and some ads. Ask ChatGPT the same question and you get three names and no links at all.
Something decided those three names. It was not an ad auction, and it was not random. Understanding how that decision gets made, and how to influence it, is what people now call GEO: Generative Engine Optimization.
If SEO was about being findable, GEO is about being quotable.
Why this suddenly matters
For twenty years, the deal was straightforward. You optimised for Google, Google sent you traffic, you converted some of it. Everyone understood the trade.
Answer engines break that deal in a specific way: they answer the question instead of routing the customer to you. The user gets what they needed and never sees a results page. If your business is one of the names in the answer, that is enormously valuable. If it is not, you are invisible. And critically, you cannot tell that you are invisible, because there is no ranking report showing you at position 47.
That is the uncomfortable part. With SEO you can at least see that you are losing. With GEO, you get silence.
How answer engines actually pick names
They are not doing anything mystical. Broadly, three things feed the answer:
1. What the model absorbed during training. If your business was described clearly on pages that existed when the model was trained, some of that is baked in. You cannot change this retroactively, which is exactly why starting now matters more than starting perfectly.
2. What it can retrieve live. Most assistants now search the web mid-answer. This is the part you can influence quickly. They pull a handful of pages and synthesise. Being in that handful is the game.
3. What it can parse confidently. This is the one everybody underestimates. A model synthesising an answer prefers sources it can extract facts from without guessing. Clear, factual, well-structured pages get quoted. Pages that bury the answer in marketing language get skipped, even when they rank well on Google.
The uncomfortable implication
A page can rank first on Google and never get quoted by an AI.
We see this constantly. The page that ranks is often a long, keyword-optimised piece that circles the topic for 2,000 words. The page that gets quoted is the one that answers the question in the first paragraph and states facts plainly.
Optimising for one does not automatically optimise for the other. They overlap, but they are not the same job.
What actually moves the needle
Answer the question in the first sentence
Not in paragraph six after the preamble. If someone asks "how much does a POS integration cost," the page should say a number, or a range, or the honest reason there is not one. Immediately.
Models extract the direct answer. Give them one to extract.
State facts, not adjectives
"We're a leading provider of innovative solutions" contains no extractable facts. There is nothing in that sentence a model can quote to a user.
"We build POS-connected ordering systems for restaurant groups with two to ten locations, in four weeks, for a fixed price quoted against a written specification" is dense with facts. Every clause is quotable.
Write the second kind.
Use structured data properly
Organization, LocalBusiness, Service, FAQPage, Product. This is not new advice, but it matters more now than it did. Structured data is a machine-readable statement of what you are and what you do, and retrieval systems lean on it heavily because it removes ambiguity.
If you run multiple locations, each one needs its own properly marked-up entity. One listing covering four restaurants is four missed chances.
Be consistent everywhere
Your name, address, phone number, and description should be identical across your site, Google Business Profile, directories, and social profiles. Models weight consistency as a confidence signal. Contradictory information makes you a risky thing to recommend, and a system that is unsure will simply name someone else.
Get mentioned in places that get retrieved
Industry publications, local press, trade directories, community sites. Not for the backlink. For the mention. Retrieval systems pull from these sources, and being described accurately on a page that gets retrieved is worth more than a hundred low-quality links.
Write the questions people actually ask
Answer engines are fed questions in natural language. "How do I connect Toast to my online ordering" is a real query. "Toast POS integration services" is a keyword phrase nobody says out loud.
Build pages around the first kind.
How to check where you stand
You cannot do this from a rank tracker. You have to ask.
Open ChatGPT, Perplexity, Gemini and Google's AI Overviews. Ask the ten questions a customer would genuinely ask before buying what you sell. Write down who gets named.
Do it monthly. That list is your ranking report. It is the only one that exists for this channel right now, and building it yourself is a genuine advantage while most of your competitors have not thought to.
The part nobody likes hearing
GEO compounds slowly and you cannot buy your way in. There is no ad product. There is no agency with a back door. The businesses that get named in twelve months are the ones being described clearly and consistently across the web today.
Which is the same reason it is worth starting now. The field is unusually open. Most businesses in most sectors have done nothing at all here. That will not be true for long.
We do this as a service. If you want to know where you currently stand, tell us your business and your market and we will run the questions and send you what the assistants actually say, at no charge and no obligation. Get in touch.