A customer asks ChatGPT where to take a client for lunch in Karen. The model names three restaurants and stops. The customer books one of the three. The other forty restaurants in the same neighbourhood never entered the conversation, and their owners have no idea the conversation happened. That gap, between ranking and being cited, is the whole problem.
Why can't AI search find your business?
A business is invisible to AI search when the machines cannot find a clear, attributable answer about it on a source they trust. Fixing that comes down to four moves: make your site machine-readable, answer the exact question your buyer asks, build consensus about your business across independent sources, and keep all of it current. Do those four and you move from absent to cited, usually inside a quarter.
Why do ranking and citation split apart?
Traditional search wants to send a user to your website, so it returns a list of links and lets the person choose. An AI engine wants to read your website so the user never has to leave the chat, so it returns an answer and names a few sources. For Google you optimise the page. For ChatGPT, Perplexity and Google AI Overviews you optimise the passage and the consensus around it.
A page can rank well and never get cited. A page can get cited without ranking in the top ten at all. Ahrefs' 2026 study found 62% of AI citations come from outside the top 10 organic results. Citation is its own discipline with its own signals, and you can be cited before you rank.
Move one: is your site machine-readable?
AI engines work by retrieval. When someone asks a question, the model runs a silent search, reads the best sources it can parse, and assembles an answer. If your homepage does not declare what your business is in structured data, the model has nothing clean to extract and defaults to whichever competitor made its data legible.
The fix is lean JSON-LD: Organization, LocalBusiness where relevant, and Article with a named author and dates. A clear heading structure the crawler can follow. This is the technical floor. If a crawler cannot reach or parse the page, nothing else on this list matters.
Move two: do your pages answer the exact question?
AI models prefer a passage that resolves a question completely, in plain language, near the top of the section. Buried answers lose to direct ones. List the ten questions your buyers actually ask an AI about your category, write the single clearest answer to each, and put that answer in the first paragraph of the page.
Explanation and nuance come after. The model reads top down and rewards the page that answers first. Open each section with a 40 to 60 word direct answer, the way this article does, because that opening block is the part the model lifts.
Move three: do independent sources agree about you?
A claim on your own website is just a claim. The same fact repeated across several independent, credible sources becomes a consensus, and consensus is what AI engines treat as reliable enough to cite. This is the move most businesses skip, because it cannot be done on your own site.
It requires your business to be mentioned elsewhere: a published article, a reputable local directory, an expert quote in a journalist's story, genuine participation in the question-and-answer communities where your buyers ask things. For a Nairobi business the legitimate starting set is small and specific: a complete Google Business Profile, one or two established local directories, and a habit of being quoted by journalists who cover your sector. Fifty low-grade directory listings do nothing. Three credible mentions of the same accurate business details do a great deal.
Move four: is any of it current?
Models favour fresh information. Date your content, update your numbers when they change, and revisit your priority pages on a schedule rather than publishing once and walking away. A page that was accurate two years ago and has not moved since reads to a model as stale, and stale loses to current when the engine picks which source to cite.
What does the sequence look like in practice?
The order is the strategy. Fix the technical floor first, because nothing downstream works without it. Then write the direct-answer pages, because they are what the engine quotes. Then earn the third-party mentions, because they are what makes the engine trust the quote. Then maintain all of it, because the engine keeps choosing the freshest reliable source. A business that does this does not need a high domain rating to start appearing in AI answers. It needs to be legible, direct, corroborated and current.
What is the one test to run this week?
Open ChatGPT and type the question your best customer would ask before choosing a business like yours. See whose name comes up. If it is not yours, you now know which of the four moves you are missing, and the order to fix them in. That single query is the cheapest diagnostic in this whole discipline.