How We Write

Practical, regulatory or platform content — how to register a business, what a feature does, a VAT threshold — is bylined African Stores Editorial Team. Heritage, culture and personal testimony is signed by a named writer with their own author page. We currently have one: Marvelous Eze. We never invent a name for a byline, a reviewer credit or a quote. If nobody real reviewed a piece, it carries no reviewer credit at all, rather than one filled in for the sake of it.

A quick answer is written for the machine: search and AI traffic, answer first, self-contained. It is not anyone's opening line and it is never dressed up as one. Key takeaways are warmer, closer to somebody leaning across the table to tell you what actually matters. FAQ answers are written the way you would explain something to your own grandchild. You should be able to tell the three apart without reading a word.

Could a general-purpose model have written this from common knowledge, with no research of its own? If yes, we do not publish it as is. A guide to starting a shop should carry something nobody else has: a real number we counted ourselves, a cost most guides get wrong, a mistake we have actually seen close a shop. “7 Tips For X” is the shape we are trying not to write.

Cultural and culinary knowledge does not need a citation — it comes from the people who live it. But anything a reader could act on has to be true and linked: law, tax, dates, money. If we say a threshold, a deadline or a fee, there is a gov.uk link under it, and it says what it actually says. We got the VAT threshold point right specifically because most guides get it wrong, and we would rather be the one that does not.

No invented metrics. If we do not have a number, we say so rather than estimate it into an average. No invented findings ahead of publication — when we say a report is coming, the numbers in the meantime are what we actually have, not what we expect to find. No invented view counts, no invented "popular articles" list assembled by hand and presented as data. No fabricated authors, reviewers or testimony.

Some articles carry a real clean markdown version at their own URL, meant for a model to read if a reader hands it over. That is not the same as fabricating a fact to make an answer sound complete, and it is not llms.txt, which the major search engines do not use. It exists because people already paste our pages into a model badly, nav and cookie banner included; this way the model reads our actual words.