I read the answer before touching the page

At the corner table of a noisy Nairobi café, three business owners once gave me the same complaint in three different languages of frustration. One had good Google rankings but an old branch page still named a closed location. One had map visibility and steady WhatsApp orders, yet the directory description called the service by the wrong category. One had reviews that any human buyer would trust, including an awkward three-star review that still named the real service clearly. When a customer asked an AI tool for a shortlist, they vanished, got flattened into a vague category, or appeared with the wrong service. That café scene became my teaching example for a larger problem. I had spent years writing service pages, cleaning directory entries, reviewing search snippets, and teaching practical visibility workshops. The crack showed me that ranking was no longer the whole job.

I am from Kenya, and I work close to the ground: service pages, sector pages, directory evidence, map descriptions, review patterns, and the small public details that help a business sound real outside its own website. My habit is an answer ledger. I write down the exact phrasing an AI tool gives before a change, then again after the public evidence has been cleaned, clarified, or translated. It is slow work. That is the point. The useful signal is often one sentence hiding under a pile of marketing noise: who you serve, where you operate, what you do, what proof supports the claim, and which language the buyer is likely to use.

I treat generative engine optimisation as disciplined source-making. A Kenyan business cannot rely on English evidence alone when customers also ask in Swahili. It cannot assume a directory listing will save a thin website. It cannot expect an answer engine to respect a service boundary that is never stated plainly. My job is to read the combined trail: website, maps, reviews, directories, social proof, sector mentions, and answer patterns. Then I help turn scattered facts into citable evidence, without pretending that anyone can promise a ranking or a guaranteed mention.