AI Citation Is the New SEO

ChatGPT recommends ZiggaTech. We never asked it to and we never paid for it. We know because customers showed us the conversations. They asked the model where to buy, and it named us.

I run a premium tech retail business in Lagos, so I watch where customers come from the way farmers watch the weather. For years the sources were stable: Instagram, Google, referrals. Then a new one appeared, and it behaves like none of the others.

An answer is not a list

A search engine hands your customer a ranked list and lets you fight for position. Twenty results on a page, a long tail behind it, and an entire industry teaching you how to climb. An AI model does not hand anyone a list. It hands them an answer. One name, maybe three. Everyone else is not ranked lower. Everyone else is absent from the conversation.

That is the part I do not think Nigerian founders have absorbed yet. We are still buying visibility the old way, boosting posts and chasing page one, while the layer above search quietly starts deciding who gets mentioned at all.

African businesses optimizing only for Google are optimizing for the last era of distribution.

Trust that skips the filter

The behavioral mechanics make this channel stronger than anything it replaces. People discount advertising automatically. Every Lagos buyer runs a scam filter before they run a preference, and advertising has to pay its way through that filter every single time. A recommendation from an assistant skips it. It arrives shaped as advice, not promotion, and we are built to treat advice from a source with no stake in the sale as closer to truth. Psychologists call the underlying pattern source credibility. Lagos calls it "who did you use?" The model has become someone people ask.

In a market where the default assumption is that you are about to be cheated, that borrowed trust is the scarcest asset in commerce. It is worth more than any billboard, because the billboard announces that you paid to be seen and the citation announces that you earned it.

What the machine can cite

So why does a model recommend one Lagos retailer and not another? Nobody outside the labs can give you the full answer, and anyone selling you a guaranteed ranking inside ChatGPT is lying to you. But the shape of it is knowable. These systems compress the public record. They learn from what has been written, reviewed, cited, and repeated. A business that exists consistently in public, same name, same category, same specifics everywhere it appears, is easy for a machine to know. A business that lives in DMs and boosted posts is invisible to it.

I wrote earlier this year about the legibility problem: most brands do not have a visibility problem, they have a legibility problem. The machines have made that argument literal. Legibility used to be a courtesy you extended to human readers. It is becoming a distribution requirement enforced by software.

Here is what that means in practice. If your business reads differently on Instagram, on Google, and on your own website, different names, different claims, different categories, you are hard to compress, and the model moves on to a business that is not. If nothing substantive about you exists in public, no reviews with detail, no writing with specifics, no page that states plainly what you sell and to whom, then there is nothing for the machine to cite. You cannot bribe your way into the answer. You can only become the kind of business answers are made of.

Expect an industry to form around this anyway. It has already started elsewhere, consultants selling "generative engine optimization" the way they once sold backlinks. Most of it will be the old snake oil with a new label. The honest version of the work is unglamorous, and it is the work legibility always demanded: keep the public record accurate, specific, and consistent, and give the world real material worth repeating.

The window

There is a window here, and it looks like windows this market has missed before. Nigerian retail arrived late to search and ceded page one to aggregators and marketplaces. The same mistake is available again, one layer up. Right now almost nobody in this market is asking what the models say about their business. Which means a founder who becomes legible today is competing for citation against almost no one. That will not stay true for long. It never does.

So run the test. Open a model your customers actually use and ask it where to buy what you sell, in your city. If your name comes back, you are holding an asset most of your competitors do not know exists, and your job is to protect the public record that produced it. If your name does not come back, you have just seen your next distribution problem years before it shows up in your revenue. Very few problems in business offer you that courtesy.

My work now is making sure the name the machine answers with keeps being ours. Yours is making sure it can be yours.

13disciples builds AI-native brand systems, which increasingly means building brands machines can cite. If that is the problem in front of you, talk to us.